BioBridge-PrimeKG (IBD) Multimodal Data Dump

In this tutorial, we will prepare Milvus database for storing and searching nodes and edges of a graph.

In particular, we are using PrimeKG multimodal data from the BioBridge project.

In [3]:

# Load necessary libraries
import os
import glob
import hydra
import cudf
import cupy as cp
import numpy as np
from pymilvus import (
    db,
    connections,
    FieldSchema,
    CollectionSchema,
    DataType,
    Collection,
    utility,
    MilvusClient
)
from tqdm import tqdm
import time
import pickle
from langchain_openai import OpenAIEmbeddings

# Load necessary libraries import os import glob import hydra import cudf import cupy as cp import numpy as np from pymilvus import ( db, connections, FieldSchema, CollectionSchema, DataType, Collection, utility, MilvusClient ) from tqdm import tqdm import time import pickle from langchain_openai import OpenAIEmbeddings

Setup OpenAI API Key for Re-Embedding¶

In [ ]:

os.environ["OPENAI_API_KEY"] = "your_openai_api_key_here"

emb_model = OpenAIEmbeddings(model="text-embedding-ada-002",
                             openai_api_key=os.environ["OPENAI_API_KEY"])

os.environ["OPENAI_API_KEY"] = "your_openai_api_key_here" emb_model = OpenAIEmbeddings(model="text-embedding-ada-002", openai_api_key=os.environ["OPENAI_API_KEY"])

Loading IBD BioBridge-PrimeKG Multimodal Data¶

First, we need to get the path to the directory containing the parquet files of nodes and edges.

For nodes and edges, we have a separate folder that contains its enrichment and embeddings.

In [6]:

# Load pickle of the graph data
with open('../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal_pyg_graph.pkl', 'rb') as f:
    graph = pickle.load(f)

# Load pickle of the graph data with open('../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal_pyg_graph.pkl', 'rb') as f: graph = pickle.load(f)

/home/awmulyadi/Repositories/AIAgents4Pharma/venv/lib/python3.12/site-packages/tqdm/auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html
  from .autonotebook import tqdm as notebook_tqdm

In [7]:

def normalize_matrix(m, axis=1):
    """
    Normalize each row of a 2D matrix using CuPy.
    
    Parameters:
    m (cupy.ndarray): 2D matrix to normalize.
    
    Returns:
    cupy.ndarray: Normalized matrix.
    """
    norms = cp.linalg.norm(m, axis=axis, keepdims=True)
    return m / norms

def normalize_vector(v):
    """
    Normalize a vector using CuPy.
    
    Parameters:
    v (cupy.ndarray): Vector to normalize.
    
    Returns:
    cupy.ndarray: Normalized vector.
    """
    v = cp.asarray(v)
    norm = cp.linalg.norm(v)
    return v / norm

def normalize_matrix(m, axis=1): """ Normalize each row of a 2D matrix using CuPy. Parameters: m (cupy.ndarray): 2D matrix to normalize. Returns: cupy.ndarray: Normalized matrix. """ norms = cp.linalg.norm(m, axis=axis, keepdims=True) return m / norms def normalize_vector(v): """ Normalize a vector using CuPy. Parameters: v (cupy.ndarray): Vector to normalize. Returns: cupy.ndarray: Normalized vector. """ v = cp.asarray(v) norm = cp.linalg.norm(v) return v / norm

Nodes Preprocessing (including re-embedding)¶

In [8]:

# Convert the list of embeddings to a 2D CuPy array (N x D)
graph_desc_x_cp = cp.asarray(graph['desc_x'].tolist())

# Normalize all rows (vectors) using broadcasting
graph_desc_x_normalized = normalize_matrix(graph_desc_x_cp, axis=1)
graph_x_normalized = [normalize_vector(v).tolist() for v in graph['x']]

# Convert the graph nodes to a cudf DataFrame
nodes_df = cudf.DataFrame({
    'node_id': graph['node_id'],
    'node_name': graph['node_name'],
    'node_type': graph['node_type'],
    'desc': graph['desc'],
    'desc_emb': graph_desc_x_normalized.tolist(),
    'feat': graph['enriched_node'],
    'feat_emb': graph_x_normalized,
})
nodes_df.reset_index(inplace=True)
nodes_df.rename(columns={'index': 'node_index'}, inplace=True)
nodes_df.head(3)

# Convert the list of embeddings to a 2D CuPy array (N x D) graph_desc_x_cp = cp.asarray(graph['desc_x'].tolist()) # Normalize all rows (vectors) using broadcasting graph_desc_x_normalized = normalize_matrix(graph_desc_x_cp, axis=1) graph_x_normalized = [normalize_vector(v).tolist() for v in graph['x']] # Convert the graph nodes to a cudf DataFrame nodes_df = cudf.DataFrame({ 'node_id': graph['node_id'], 'node_name': graph['node_name'], 'node_type': graph['node_type'], 'desc': graph['desc'], 'desc_emb': graph_desc_x_normalized.tolist(), 'feat': graph['enriched_node'], 'feat_emb': graph_x_normalized, }) nodes_df.reset_index(inplace=True) nodes_df.rename(columns={'index': 'node_index'}, inplace=True) nodes_df.head(3)

Out[8]:

	node_index	node_id	node_name	node_type	desc	desc_emb	feat	feat_emb
0	0	SMAD3_(144)	SMAD3	gene/protein	SMAD3 belongs to gene/protein node. SMAD3 is S...	[0.02974936784975063, 0.05350021171537046, -0....	MSSILPFTPPIVKRLLGWKKGEQNGQEEKWCEKAVKSLVKKLKKTG...	[-0.0010794274069904548, -0.0028632148270051, ...
1	1	IL10RB_(179)	IL10RB	gene/protein	IL10RB belongs to gene/protein node. IL10RB is...	[0.02842173040130417, 0.01986006372730412, -0....	MAWSLGSWLGGCLLVSALGMVPPPENVRMNSVNFKNILQWESPAFA...	[-0.007157766077247574, 0.006195289622587354, ...
2	2	GNA12_(192)	GNA12	gene/protein	GNA12 belongs to gene/protein node. GNA12 is G...	[0.003668847841835145, 0.051380571197126614, -...	MSGVVRTLSRCLLPAEAGGARERRAGSGARDAEREARRRSRDIDAL...	[-0.001562959383761, -0.01338132129666802, -0....

Optional: Nodes Re-embedding usin OpenAI¶

You may skip this section if you want to stick with 'nomic-embed-text' embedding model as default one.

In [9]:

# Checking embedding dimensions before proceeding
emb_dim = len(nodes_df['desc_emb'].iloc[0])
print(f"Embedding dimension: {emb_dim}")

# Checking embedding dimensions before proceeding emb_dim = len(nodes_df['desc_emb'].iloc[0]) print(f"Embedding dimension: {emb_dim}")

Embedding dimension: 768

In [11]:

# For textual data, we will re-embed the descriptions using OpenAI embeddings
mini_batch_size = 100
desc_embeddings = []
for i in tqdm(range(0, nodes_df.shape[0], mini_batch_size), desc="Re-embedding descriptions"):
    batch = nodes_df['desc'].to_pandas().tolist()[i:i+mini_batch_size]
    embeddings = emb_model.embed_documents(batch)
    desc_embeddings.extend(embeddings)
nodes_df['desc_emb'] = desc_embeddings

# For textual data, we will re-embed the descriptions using OpenAI embeddings mini_batch_size = 100 desc_embeddings = [] for i in tqdm(range(0, nodes_df.shape[0], mini_batch_size), desc="Re-embedding descriptions"): batch = nodes_df['desc'].to_pandas().tolist()[i:i+mini_batch_size] embeddings = emb_model.embed_documents(batch) desc_embeddings.extend(embeddings) nodes_df['desc_emb'] = desc_embeddings

Re-embedding descriptions: 100%|██████████| 30/30 [00:27<00:00,  1.07it/s]

In [12]:

# Checking embeddings dimensions after re-embedding
emb_dim = len(nodes_df['desc_emb'].iloc[0])
print(f"Embedding dimension after re-embedding: {emb_dim}")

# Checking embeddings dimensions after re-embedding emb_dim = len(nodes_df['desc_emb'].iloc[0]) print(f"Embedding dimension after re-embedding: {emb_dim}")

Embedding dimension after re-embedding: 1536

In [15]:

# Get the text-based nodes for re-embedding
text_based_df = nodes_df[nodes_df.node_type.isin(['disease', 'biological_process', 'cellular_component', 'molecular_function'])]

for nt, text_based_df_ in text_based_df.groupby("node_type"):
    print(f"Re-embedding {nt} nodes")
    # Checking embedding dimensions before proceeding
    emb_dim = len(text_based_df_['feat_emb'].iloc[0])
    print(f"Embedding dimension: {emb_dim}")
    print('---')

# Get the text-based nodes for re-embedding text_based_df = nodes_df[nodes_df.node_type.isin(['disease', 'biological_process', 'cellular_component', 'molecular_function'])] for nt, text_based_df_ in text_based_df.groupby("node_type"): print(f"Re-embedding {nt} nodes") # Checking embedding dimensions before proceeding emb_dim = len(text_based_df_['feat_emb'].iloc[0]) print(f"Embedding dimension: {emb_dim}") print('---')

Re-embedding biological_process nodes
Embedding dimension: 768
---
Re-embedding cellular_component nodes
Embedding dimension: 768
---
Re-embedding disease nodes
Embedding dimension: 768
---
Re-embedding molecular_function nodes
Embedding dimension: 768
---

In [17]:

# Update textual pre-loaded embeddings with OpenAI embeddings
# Since the records of nodes has large amount of data, we will split them into mini-batches
mini_batch_size = 100
text_node_indexes = []
text_node_embeddings = []
for i in tqdm(range(0, text_based_df.shape[0], mini_batch_size), desc="Re-embedding text nodes"):
    outputs = emb_model.embed_documents(text_based_df.to_pandas().feat.values.tolist()[i:i+mini_batch_size])
    text_node_indexes.extend(text_based_df.to_pandas().node_index.values.tolist()[i:i+mini_batch_size])
    text_node_embeddings.extend(outputs)
dic_text_embeddings = dict(zip(text_node_indexes, text_node_embeddings))

# Update textual pre-loaded embeddings with OpenAI embeddings # Since the records of nodes has large amount of data, we will split them into mini-batches mini_batch_size = 100 text_node_indexes = [] text_node_embeddings = [] for i in tqdm(range(0, text_based_df.shape[0], mini_batch_size), desc="Re-embedding text nodes"): outputs = emb_model.embed_documents(text_based_df.to_pandas().feat.values.tolist()[i:i+mini_batch_size]) text_node_indexes.extend(text_based_df.to_pandas().node_index.values.tolist()[i:i+mini_batch_size]) text_node_embeddings.extend(outputs) dic_text_embeddings = dict(zip(text_node_indexes, text_node_embeddings))

Re-embedding text nodes: 100%|██████████| 22/22 [00:58<00:00,  2.65s/it]

In [19]:

# Replace the embeddings of the nodes with the updated embeddings for text-based nodes
nodes_df["feat_emb"] = nodes_df.to_pandas().apply(lambda x: dic_text_embeddings[x["node_index"]] if x["node_index"] in dic_text_embeddings else x["feat_emb"], axis=1)

# Replace the embeddings of the nodes with the updated embeddings for text-based nodes nodes_df["feat_emb"] = nodes_df.to_pandas().apply(lambda x: dic_text_embeddings[x["node_index"]] if x["node_index"] in dic_text_embeddings else x["feat_emb"], axis=1)

In [20]:

# Get the text-based nodes for re-embedding
text_based_df = nodes_df[nodes_df.node_type.isin(['disease', 'biological_process', 'cellular_component', 'molecular_function'])]

for nt, text_based_df_ in text_based_df.groupby("node_type"):
    print(f"Re-embedding {nt} nodes")
    # Checking embedding dimensions before proceeding
    emb_dim = len(text_based_df_['feat_emb'].iloc[0])
    print(f"Embedding dimension: {emb_dim}")
    print('---')

# Get the text-based nodes for re-embedding text_based_df = nodes_df[nodes_df.node_type.isin(['disease', 'biological_process', 'cellular_component', 'molecular_function'])] for nt, text_based_df_ in text_based_df.groupby("node_type"): print(f"Re-embedding {nt} nodes") # Checking embedding dimensions before proceeding emb_dim = len(text_based_df_['feat_emb'].iloc[0]) print(f"Embedding dimension: {emb_dim}") print('---')

Re-embedding biological_process nodes
Embedding dimension: 1536
---
Re-embedding cellular_component nodes
Embedding dimension: 1536
---
Re-embedding disease nodes
Embedding dimension: 1536
---
Re-embedding molecular_function nodes
Embedding dimension: 1536
---

Edges Preprocessing (including re-embedding)¶

In [21]:

# Convert the list of edge embeddings to a 2D CuPy array (M x D)
graph_edge_attr_cp = cp.asarray(graph['edge_attr'].tolist())

# Normalize all rows (vectors) using broadcasting
graph_edge_attr_normalized = normalize_matrix(graph_edge_attr_cp, axis=1)

# Convert the graph edges to a cudf DataFrame
edges_df = cudf.DataFrame({
    'triplet_index': graph['triplet_index'],
    'head_id': graph['head_id'],
    'head_name': graph['head_name'],
    'tail_id': graph['tail_id'],
    'tail_name': graph['tail_name'],
    'display_relation': graph['display_relation'],
    'edge_type': graph['edge_type'],
    'edge_type_str': ['|'.join(e) for e in graph['edge_type']],
    'feat': graph['enriched_edge'],
    'edge_emb': graph_edge_attr_normalized.tolist(),
})
edges_df = edges_df.merge(
    nodes_df[['node_index', 'node_id']],
    left_on='head_id',
    right_on='node_id',
    how='left'
)
edges_df.rename(columns={'node_index': 'head_index'}, inplace=True)
edges_df.drop(columns=['node_id'], inplace=True)
edges_df = edges_df.merge(
    nodes_df[['node_index', 'node_id']],
    left_on='tail_id',
    right_on='node_id',
    how='left'
)
edges_df.rename(columns={'node_index': 'tail_index'}, inplace=True)
edges_df.drop(columns=['node_id'], inplace=True)
edges_df.head(3)

# Convert the list of edge embeddings to a 2D CuPy array (M x D) graph_edge_attr_cp = cp.asarray(graph['edge_attr'].tolist()) # Normalize all rows (vectors) using broadcasting graph_edge_attr_normalized = normalize_matrix(graph_edge_attr_cp, axis=1) # Convert the graph edges to a cudf DataFrame edges_df = cudf.DataFrame({ 'triplet_index': graph['triplet_index'], 'head_id': graph['head_id'], 'head_name': graph['head_name'], 'tail_id': graph['tail_id'], 'tail_name': graph['tail_name'], 'display_relation': graph['display_relation'], 'edge_type': graph['edge_type'], 'edge_type_str': ['|'.join(e) for e in graph['edge_type']], 'feat': graph['enriched_edge'], 'edge_emb': graph_edge_attr_normalized.tolist(), }) edges_df = edges_df.merge( nodes_df[['node_index', 'node_id']], left_on='head_id', right_on='node_id', how='left' ) edges_df.rename(columns={'node_index': 'head_index'}, inplace=True) edges_df.drop(columns=['node_id'], inplace=True) edges_df = edges_df.merge( nodes_df[['node_index', 'node_id']], left_on='tail_id', right_on='node_id', how='left' ) edges_df.rename(columns={'node_index': 'tail_index'}, inplace=True) edges_df.drop(columns=['node_id'], inplace=True) edges_df.head(3)

Out[21]:

	triplet_index	head_id	head_name	tail_id	tail_name	display_relation	edge_type	edge_type_str	feat	edge_emb	head_index	tail_index
0	8602	cytokine-mediated signaling pathway_(47242)	cytokine-mediated signaling pathway	IL10RB_(179)	IL10RB	interacts with	[biological_process, interacts with, gene/prot...	biological_process|interacts with|gene/protein	cytokine-mediated signaling pathway (biologica...	[0.016838406414606846, 0.019238545922865967, -...	1455	1
1	8603	cytokine-mediated signaling pathway_(47242)	cytokine-mediated signaling pathway	IL12B_(6168)	IL12B	interacts with	[biological_process, interacts with, gene/prot...	biological_process|interacts with|gene/protein	cytokine-mediated signaling pathway (biologica...	[0.018197947379867397, 0.03141968316046658, -0...	1455	59
2	8604	cytokine-mediated signaling pathway_(47242)	cytokine-mediated signaling pathway	IRF5_(3646)	IRF5	interacts with	[biological_process, interacts with, gene/prot...	biological_process|interacts with|gene/protein	cytokine-mediated signaling pathway (biologica...	[0.018029207941198132, 0.019414354880667273, -...	1455	46

Optional: Edges Re-embedding using OpenAI¶

In [22]:

# Checking embeddings dimensions after re-embedding
emb_dim = len(edges_df['edge_emb'].iloc[0])
print(f"Embedding dimension after re-embedding: {emb_dim}")

# Checking embeddings dimensions after re-embedding emb_dim = len(edges_df['edge_emb'].iloc[0]) print(f"Embedding dimension after re-embedding: {emb_dim}")

Embedding dimension after re-embedding: 768

In [23]:

# For textual data, we will re-embed the descriptions using OpenAI embeddings
mini_batch_size = 100
edge_embeddings = []
for i in tqdm(range(0, edges_df.shape[0], mini_batch_size), desc="Re-embedding edges"):
    batch = edges_df['feat'].to_pandas().tolist()[i:i+mini_batch_size]
    embeddings = emb_model.embed_documents(batch)
    edge_embeddings.extend(embeddings)
edges_df['edge_emb'] = edge_embeddings

# For textual data, we will re-embed the descriptions using OpenAI embeddings mini_batch_size = 100 edge_embeddings = [] for i in tqdm(range(0, edges_df.shape[0], mini_batch_size), desc="Re-embedding edges"): batch = edges_df['feat'].to_pandas().tolist()[i:i+mini_batch_size] embeddings = emb_model.embed_documents(batch) edge_embeddings.extend(embeddings) edges_df['edge_emb'] = edge_embeddings

Re-embedding edges: 100%|██████████| 113/113 [01:37<00:00,  1.16it/s]

In [24]:

# Checking embeddings dimensions after re-embedding
emb_dim = len(edges_df['edge_emb'].iloc[0])
print(f"Embedding dimension after re-embedding: {emb_dim}")

# Checking embeddings dimensions after re-embedding emb_dim = len(edges_df['edge_emb'].iloc[0]) print(f"Embedding dimension after re-embedding: {emb_dim}")

Embedding dimension after re-embedding: 1536

Storing dataframes¶

In [25]:

# Store the DataFrame into compressed parquet files
storage_path = "../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/"
os.makedirs(storage_path, exist_ok=True)

# Nodes enrichment & embeddings
nodes_enrichment = nodes_df[['node_index', 'node_id', 'node_name', 'node_type', 'desc', 'feat', ]].to_pandas().copy()
os.makedirs(os.path.join(storage_path, 'nodes', 'enrichment'), exist_ok=True)
for nt, nodes_df_ in nodes_enrichment.groupby('node_type'):
    print(nt, nodes_df_.shape)
    nodes_df_.to_parquet(os.path.join(storage_path, 'nodes', 'enrichment', f'{nt.replace('/', '_')}.parquet.gzip'), 
                         compression='gzip',
                         index=False)
print("Nodes enrichment saved.")
print('---')

nodes_embeddings = nodes_df[['node_index', 'node_id', 'node_type', 'desc_emb', 'feat_emb']].to_pandas().copy()
os.makedirs(os.path.join(storage_path, 'nodes', 'embedding'), exist_ok=True)
for nt, nodes_df_ in nodes_embeddings.groupby('node_type'):
    print(nt, nodes_df_.shape)
    nodes_df_[['node_index', 'node_id', 'desc_emb', 'feat_emb']].to_parquet(os.path.join(storage_path, 'nodes', 'embedding', f'{nt.replace('/', '_')}.parquet.gzip'), 
                                                                 compression='gzip',
                                                                 index=False)
print("Nodes embeddings saved.")

# Store the DataFrame into compressed parquet files storage_path = "../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/" os.makedirs(storage_path, exist_ok=True) # Nodes enrichment & embeddings nodes_enrichment = nodes_df[['node_index', 'node_id', 'node_name', 'node_type', 'desc', 'feat', ]].to_pandas().copy() os.makedirs(os.path.join(storage_path, 'nodes', 'enrichment'), exist_ok=True) for nt, nodes_df_ in nodes_enrichment.groupby('node_type'): print(nt, nodes_df_.shape) nodes_df_.to_parquet(os.path.join(storage_path, 'nodes', 'enrichment', f'{nt.replace('/', '_')}.parquet.gzip'), compression='gzip', index=False) print("Nodes enrichment saved.") print('---') nodes_embeddings = nodes_df[['node_index', 'node_id', 'node_type', 'desc_emb', 'feat_emb']].to_pandas().copy() os.makedirs(os.path.join(storage_path, 'nodes', 'embedding'), exist_ok=True) for nt, nodes_df_ in nodes_embeddings.groupby('node_type'): print(nt, nodes_df_.shape) nodes_df_[['node_index', 'node_id', 'desc_emb', 'feat_emb']].to_parquet(os.path.join(storage_path, 'nodes', 'embedding', f'{nt.replace('/', '_')}.parquet.gzip'), compression='gzip', index=False) print("Nodes embeddings saved.")

biological_process (1615, 6)
cellular_component (202, 6)
disease (7, 6)
drug (748, 6)
gene/protein (102, 6)
molecular_function (317, 6)
Nodes enrichment saved.
---
biological_process (1615, 5)
cellular_component (202, 5)
disease (7, 5)
drug (748, 5)
gene/protein (102, 5)
molecular_function (317, 5)
Nodes embeddings saved.

In [26]:

# Edges enrichment & embeddings
edges_enrichment = edges_df[['triplet_index', 'head_id', 'head_index', 'tail_id', 'tail_index', 
                             'edge_type', 'edge_type_str', 'display_relation', 'feat']].to_pandas().copy()
os.makedirs(os.path.join(storage_path, 'edges', 'enrichment'), exist_ok=True)
edges_enrichment.to_parquet(os.path.join(storage_path, 'edges', 'enrichment', f'edges.parquet.gzip'), 
                            compression='gzip',
                            index=False)
print("Edges enrichment saved.")
print('---')

edges_embeddings = edges_df[['triplet_index', 'head_index', 'tail_index', 'edge_emb']].to_pandas().copy()
os.makedirs(os.path.join(storage_path, 'edges', 'embedding'), exist_ok=True)
chunk_size = 1000
for i in range(0, edges_embeddings.shape[0], chunk_size):
    et = f'edges_{i // chunk_size}'
    edges_embeddings_chunk = edges_embeddings.iloc[i:i + chunk_size]
    # Save each chunk to a separate parquet file
    edges_embeddings_chunk.to_parquet(os.path.join(storage_path, 'edges', 'embedding', f'{et}.parquet.gzip'), 
                                      compression='gzip',
                                      index=False)
print("Edges embeddings saved.")

# Edges enrichment & embeddings edges_enrichment = edges_df[['triplet_index', 'head_id', 'head_index', 'tail_id', 'tail_index', 'edge_type', 'edge_type_str', 'display_relation', 'feat']].to_pandas().copy() os.makedirs(os.path.join(storage_path, 'edges', 'enrichment'), exist_ok=True) edges_enrichment.to_parquet(os.path.join(storage_path, 'edges', 'enrichment', f'edges.parquet.gzip'), compression='gzip', index=False) print("Edges enrichment saved.") print('---') edges_embeddings = edges_df[['triplet_index', 'head_index', 'tail_index', 'edge_emb']].to_pandas().copy() os.makedirs(os.path.join(storage_path, 'edges', 'embedding'), exist_ok=True) chunk_size = 1000 for i in range(0, edges_embeddings.shape[0], chunk_size): et = f'edges_{i // chunk_size}' edges_embeddings_chunk = edges_embeddings.iloc[i:i + chunk_size] # Save each chunk to a separate parquet file edges_embeddings_chunk.to_parquet(os.path.join(storage_path, 'edges', 'embedding', f'{et}.parquet.gzip'), compression='gzip', index=False) print("Edges embeddings saved.")

Edges enrichment saved.
---
Edges embeddings saved.

Loading dataframes¶

In [27]:

# Set storage path for the dataframes
storage_path = "../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/"

# Set storage path for the dataframes storage_path = "../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/"

In [28]:

# Loop over nodes and edges
graph_dict = {}
for element in ["nodes", "edges"]:
    # Make an empty dictionary for each folder
    graph_dict[element] = {}
    for stage in ["enrichment", "embedding"]:
        print(element, stage)
        # Create the file pattern for the current subfolder
        file_list = glob.glob(os.path.join(storage_path,
                                           element,
                                           stage, '*.parquet.gzip'))
        print(file_list)
        # Read and concatenate all dataframes in the folder
        # Except the edges embedding, which is too large to read in one go
        # We are using a chunk size to read the edges embedding in smaller parts instead
        if element == "edges" and stage == "embedding":
            # For edges embedding, only read two columns: triplet_index and edge_emb
            # graph_dict[element][stage] = cudf.concat([cudf.read_parquet(f, columns=["triplet_index", "edge_emb"]) for f in file_list[:2]], ignore_index=True)
            # Loop by chunks
            # file_list = file_list[:2]
            chunk_size = 5
            graph_dict[element][stage] = []
            for i in range(0, len(file_list), chunk_size):
                chunk_files = file_list[i:i+chunk_size]
                chunk_df = cudf.concat([cudf.read_parquet(f, columns=["triplet_index", "edge_emb"]) for f in chunk_files], ignore_index=True)
                graph_dict[element][stage].append(chunk_df)
        else:
            # For nodes and edges enrichment, read and concatenate all dataframes in the folder
            # This includes the nodes embedding, which is small enough to read in one go
            graph_dict[element][stage] = cudf.concat([cudf.read_parquet(f) for f in file_list], ignore_index=True)

# Loop over nodes and edges graph_dict = {} for element in ["nodes", "edges"]: # Make an empty dictionary for each folder graph_dict[element] = {} for stage in ["enrichment", "embedding"]: print(element, stage) # Create the file pattern for the current subfolder file_list = glob.glob(os.path.join(storage_path, element, stage, '*.parquet.gzip')) print(file_list) # Read and concatenate all dataframes in the folder # Except the edges embedding, which is too large to read in one go # We are using a chunk size to read the edges embedding in smaller parts instead if element == "edges" and stage == "embedding": # For edges embedding, only read two columns: triplet_index and edge_emb # graph_dict[element][stage] = cudf.concat([cudf.read_parquet(f, columns=["triplet_index", "edge_emb"]) for f in file_list[:2]], ignore_index=True) # Loop by chunks # file_list = file_list[:2] chunk_size = 5 graph_dict[element][stage] = [] for i in range(0, len(file_list), chunk_size): chunk_files = file_list[i:i+chunk_size] chunk_df = cudf.concat([cudf.read_parquet(f, columns=["triplet_index", "edge_emb"]) for f in chunk_files], ignore_index=True) graph_dict[element][stage].append(chunk_df) else: # For nodes and edges enrichment, read and concatenate all dataframes in the folder # This includes the nodes embedding, which is small enough to read in one go graph_dict[element][stage] = cudf.concat([cudf.read_parquet(f) for f in file_list], ignore_index=True)

nodes enrichment
['../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/nodes/enrichment/gene_protein.parquet.gzip', '../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/nodes/enrichment/disease.parquet.gzip', '../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/nodes/enrichment/biological_process.parquet.gzip', '../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/nodes/enrichment/cellular_component.parquet.gzip', '../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/nodes/enrichment/molecular_function.parquet.gzip', '../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/nodes/enrichment/drug.parquet.gzip']
nodes embedding
['../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/nodes/embedding/gene_protein.parquet.gzip', '../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/nodes/embedding/disease.parquet.gzip', '../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/nodes/embedding/biological_process.parquet.gzip', '../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/nodes/embedding/cellular_component.parquet.gzip', '../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/nodes/embedding/molecular_function.parquet.gzip', '../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/nodes/embedding/drug.parquet.gzip']
edges enrichment
['../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/edges/enrichment/edges.parquet.gzip']
edges embedding
['../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/edges/embedding/edges_3.parquet.gzip', '../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/edges/embedding/edges_5.parquet.gzip', '../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/edges/embedding/edges_0.parquet.gzip', '../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/edges/embedding/edges_9.parquet.gzip', '../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/edges/embedding/edges_8.parquet.gzip', '../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/edges/embedding/edges_7.parquet.gzip', '../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/edges/embedding/edges_11.parquet.gzip', '../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/edges/embedding/edges_1.parquet.gzip', '../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/edges/embedding/edges_2.parquet.gzip', '../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/edges/embedding/edges_4.parquet.gzip', '../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/edges/embedding/edges_6.parquet.gzip', '../../../aiagents4pharma/talk2knowledgegraphs/tests/files/biobridge_multimodal/edges/embedding/edges_10.parquet.gzip']

In [29]:

# Get nodes enrichment and embedding dataframes
nodes_enrichment_df = graph_dict['nodes']['enrichment']
nodes_embedding_df = graph_dict['nodes']['embedding']

# Get edges enrichment and embedding dataframes
edges_enrichment_df = graph_dict['edges']['enrichment']
edges_embedding_df = graph_dict['edges']['embedding'] # !!consisted of a list of dataframes!!

# Get nodes enrichment and embedding dataframes nodes_enrichment_df = graph_dict['nodes']['enrichment'] nodes_embedding_df = graph_dict['nodes']['embedding'] # Get edges enrichment and embedding dataframes edges_enrichment_df = graph_dict['edges']['enrichment'] edges_embedding_df = graph_dict['edges']['embedding'] # !!consisted of a list of dataframes!!

In [45]:

# Merge nodes enrichment and embedding dataframes
merged_nodes_df = nodes_enrichment_df.merge(
    nodes_embedding_df[["node_id", "desc_emb", "feat_emb"]],
    on="node_id",
    how="left"
)
# del nodes_enrichment_df, nodes_embedding_df  # Free memory

# Merge nodes enrichment and embedding dataframes merged_nodes_df = nodes_enrichment_df.merge( nodes_embedding_df[["node_id", "desc_emb", "feat_emb"]], on="node_id", how="left" ) # del nodes_enrichment_df, nodes_embedding_df # Free memory

In [30]:

# Check dataframe
nodes_enrichment_df.head(3)

# Check dataframe nodes_enrichment_df.head(3)

Out[30]:

	node_index	node_id	node_name	node_type	desc	feat
0	0	SMAD3_(144)	SMAD3	gene/protein	SMAD3 belongs to gene/protein node. SMAD3 is S...	MSSILPFTPPIVKRLLGWKKGEQNGQEEKWCEKAVKSLVKKLKKTG...
1	1	IL10RB_(179)	IL10RB	gene/protein	IL10RB belongs to gene/protein node. IL10RB is...	MAWSLGSWLGGCLLVSALGMVPPPENVRMNSVNFKNILQWESPAFA...
2	2	GNA12_(192)	GNA12	gene/protein	GNA12 belongs to gene/protein node. GNA12 is G...	MSGVVRTLSRCLLPAEAGGARERRAGSGARDAEREARRRSRDIDAL...

In [31]:

# Check dataframe
nodes_embedding_df.head(3)

# Check dataframe nodes_embedding_df.head(3)

Out[31]:

	node_index	node_id	desc_emb	feat_emb
0	0	SMAD3_(144)	[-0.03699171170592308, -0.005479035433381796, ...	[-0.0010794274069904548, -0.0028632148270051, ...
1	1	IL10RB_(179)	[-0.02927332930266857, -0.0068625640124082565,...	[-0.007157766077247574, 0.006195289622587354, ...
2	2	GNA12_(192)	[-0.02188265137374401, -0.01718498021364212, -...	[-0.001562959383761, -0.01338132129666802, -0....

In [32]:

# Check dataframe
edges_enrichment_df.head(3)

# Check dataframe edges_enrichment_df.head(3)

Out[32]:

	triplet_index	head_id	head_index	tail_id	tail_index	edge_type	edge_type_str	display_relation	feat
0	8602	cytokine-mediated signaling pathway_(47242)	1455	IL10RB_(179)	1	[biological_process, interacts with, gene/prot...	biological_process|interacts with|gene/protein	interacts with	cytokine-mediated signaling pathway (biologica...
1	8603	cytokine-mediated signaling pathway_(47242)	1455	IL12B_(6168)	59	[biological_process, interacts with, gene/prot...	biological_process|interacts with|gene/protein	interacts with	cytokine-mediated signaling pathway (biologica...
2	8604	cytokine-mediated signaling pathway_(47242)	1455	IRF5_(3646)	46	[biological_process, interacts with, gene/prot...	biological_process|interacts with|gene/protein	interacts with	cytokine-mediated signaling pathway (biologica...

In [33]:

# Check dataframes
len(edges_embedding_df)

# Check dataframes len(edges_embedding_df)

Out[33]:

3

In [34]:

# Check the first chunk of edges embedding
edges_embedding_df[0].head(3)

# Check the first chunk of edges embedding edges_embedding_df[0].head(3)

Out[34]:

	triplet_index	edge_emb
0	7408	[-0.030603667721152306, -0.020534928888082504,...
1	7071	[-0.028667431324720383, -0.004508184734731913,...
2	7072	[-0.04058527573943138, -0.010810465551912785, ...

Setup Milvus Database¶

In [91]:

# Configuration for Milvus
milvus_host = "localhost"
milvus_port = "19530"
milvus_uri = "http://localhost:19530"
milvus_token = "root:Milvus"
milvus_user = "root"
milvus_password = "Milvus"
milvus_database = "t2kg_primekg"

# Configuration for Milvus milvus_host = "localhost" milvus_port = "19530" milvus_uri = "http://localhost:19530" milvus_token = "root:Milvus" milvus_user = "root" milvus_password = "Milvus" milvus_database = "t2kg_primekg"

In [ ]:

# Connect to Milvus
connections.connect(
    alias="default",
    host=milvus_host,
    port=milvus_port,
    user=milvus_user,
    password=milvus_password
)

# Connect to Milvus connections.connect( alias="default", host=milvus_host, port=milvus_port, user=milvus_user, password=milvus_password )

In [42]:

# Check if the database exists, create if it doesn't
if milvus_database not in db.list_database():
    db.create_database(milvus_database)

# Switch to the desired database
db.using_database(milvus_database)

# Check if the database exists, create if it doesn't if milvus_database not in db.list_database(): db.create_database(milvus_database) # Switch to the desired database db.using_database(milvus_database)

In [43]:

# List all collections
for coll in utility.list_collections():
    print(f"Collection: {coll}")

    # Load the collection to get stats
    collection = Collection(name=coll)
    print(collection.num_entities)
    
    # Drop the collection if it exists
    if utility.has_collection(coll):
        print(f"Dropping collection: {coll}")
        utility.drop_collection(coll)

# List all collections for coll in utility.list_collections(): print(f"Collection: {coll}") # Load the collection to get stats collection = Collection(name=coll) print(collection.num_entities) # Drop the collection if it exists if utility.has_collection(coll): print(f"Dropping collection: {coll}") utility.drop_collection(coll)

In [44]:

# A helper function to chunk the data into smaller parts
# Utility: chunk generator
def chunked(data_list, chunk_size):
    for i in range(0, len(data_list), chunk_size):
        yield data_list[i:i + chunk_size]

# A helper function to chunk the data into smaller parts # Utility: chunk generator def chunked(data_list, chunk_size): for i in range(0, len(data_list), chunk_size): yield data_list[i:i + chunk_size]

Building Node Collection (Description Embedding)¶

In [46]:

%%time

# Configuration for Milvus collection
node_coll_name = f"{milvus_database}_nodes"

# Define schema for the collection
# Leave out the feat and feat_emb fields for now
desc_emb_dim = len(merged_nodes_df.iloc[0]['desc_emb'].to_arrow().to_pylist()[0])
node_fields = [
    FieldSchema(name="node_index", dtype=DataType.INT64, is_primary=True),
    FieldSchema(name="node_id", dtype=DataType.VARCHAR, max_length=1024),
    FieldSchema(name="node_name", dtype=DataType.VARCHAR, max_length=1024,
                enable_analyzer=True, enable_match=True),
    FieldSchema(name="node_type", dtype=DataType.VARCHAR, max_length=1024,
                enable_analyzer=True, enable_match=True),
    FieldSchema(name="desc", dtype=DataType.VARCHAR, max_length=40960,
                enable_analyzer=True, enable_match=True),
    FieldSchema(name="desc_emb", dtype=DataType.FLOAT_VECTOR, dim=desc_emb_dim),
]
schema  = CollectionSchema(fields=node_fields, description=f"Schema for collection {node_coll_name}")

# Create collection if it doesn't exist
if not utility.has_collection(node_coll_name):
    collection = Collection(name=node_coll_name, schema=schema)
else:
    collection = Collection(name=node_coll_name)

# Create indexes
collection.create_index(
    field_name="node_index",
    index_params={"index_type": "STL_SORT"}, # STL_SORT
    index_name="node_index_index"
)
# Create index for node_name, node_type, desc fields (inverted)
collection.create_index(
    field_name="node_name",
    index_params={"index_type": "INVERTED"},
    index_name="node_name_index"
)
collection.create_index(
    field_name="node_type",
    index_params={"index_type": "INVERTED"},
    index_name="node_type_index"
)
collection.create_index(
    field_name="desc",
    index_params={"index_type": "INVERTED"},
    index_name="desc_index"
)
collection.create_index(
    field_name="desc_emb",
    index_params={"index_type": "GPU_CAGRA", "metric_type": "IP"}, # AUTOINDEX
    index_name="desc_emb_index"
)

# Prepare data for insertion
# Normalize the embeddings
graph_desc_emb_cp = cp.asarray(merged_nodes_df["desc_emb"].list.leaves).astype(cp.float32).reshape(merged_nodes_df.shape[0], -1)
graph_desc_emb_norm = normalize_matrix(graph_desc_emb_cp, axis=1)
data = [
    merged_nodes_df["node_index"].to_arrow().to_pylist(),
    merged_nodes_df["node_id"].to_arrow().to_pylist(),
    merged_nodes_df["node_name"].to_arrow().to_pylist(),
    merged_nodes_df["node_type"].to_arrow().to_pylist(),
    merged_nodes_df["desc"].to_arrow().to_pylist(),
    graph_desc_emb_norm.tolist(), # Use normalized embeddings
]

# Insert data in batches
batch_size = 500
total = len(data[0])
for i in tqdm(range(0, total, batch_size)):
    batch = [
        col[i:i+batch_size] for col in data
    ]
    collection.insert(batch)

# Flush to persist data
collection.flush()

# Get collection stats
print(collection.num_entities)

%%time # Configuration for Milvus collection node_coll_name = f"{milvus_database}_nodes" # Define schema for the collection # Leave out the feat and feat_emb fields for now desc_emb_dim = len(merged_nodes_df.iloc[0]['desc_emb'].to_arrow().to_pylist()[0]) node_fields = [ FieldSchema(name="node_index", dtype=DataType.INT64, is_primary=True), FieldSchema(name="node_id", dtype=DataType.VARCHAR, max_length=1024), FieldSchema(name="node_name", dtype=DataType.VARCHAR, max_length=1024, enable_analyzer=True, enable_match=True), FieldSchema(name="node_type", dtype=DataType.VARCHAR, max_length=1024, enable_analyzer=True, enable_match=True), FieldSchema(name="desc", dtype=DataType.VARCHAR, max_length=40960, enable_analyzer=True, enable_match=True), FieldSchema(name="desc_emb", dtype=DataType.FLOAT_VECTOR, dim=desc_emb_dim), ] schema = CollectionSchema(fields=node_fields, description=f"Schema for collection {node_coll_name}") # Create collection if it doesn't exist if not utility.has_collection(node_coll_name): collection = Collection(name=node_coll_name, schema=schema) else: collection = Collection(name=node_coll_name) # Create indexes collection.create_index( field_name="node_index", index_params={"index_type": "STL_SORT"}, # STL_SORT index_name="node_index_index" ) # Create index for node_name, node_type, desc fields (inverted) collection.create_index( field_name="node_name", index_params={"index_type": "INVERTED"}, index_name="node_name_index" ) collection.create_index( field_name="node_type", index_params={"index_type": "INVERTED"}, index_name="node_type_index" ) collection.create_index( field_name="desc", index_params={"index_type": "INVERTED"}, index_name="desc_index" ) collection.create_index( field_name="desc_emb", index_params={"index_type": "GPU_CAGRA", "metric_type": "IP"}, # AUTOINDEX index_name="desc_emb_index" ) # Prepare data for insertion # Normalize the embeddings graph_desc_emb_cp = cp.asarray(merged_nodes_df["desc_emb"].list.leaves).astype(cp.float32).reshape(merged_nodes_df.shape[0], -1) graph_desc_emb_norm = normalize_matrix(graph_desc_emb_cp, axis=1) data = [ merged_nodes_df["node_index"].to_arrow().to_pylist(), merged_nodes_df["node_id"].to_arrow().to_pylist(), merged_nodes_df["node_name"].to_arrow().to_pylist(), merged_nodes_df["node_type"].to_arrow().to_pylist(), merged_nodes_df["desc"].to_arrow().to_pylist(), graph_desc_emb_norm.tolist(), # Use normalized embeddings ] # Insert data in batches batch_size = 500 total = len(data[0]) for i in tqdm(range(0, total, batch_size)): batch = [ col[i:i+batch_size] for col in data ] collection.insert(batch) # Flush to persist data collection.flush() # Get collection stats print(collection.num_entities)

100%|██████████| 6/6 [00:04<00:00,  1.46it/s]

2991
CPU times: user 1.3 s, sys: 137 ms, total: 1.44 s
Wall time: 8.61 s

In [47]:

# List all collections
for coll in utility.list_collections():
    print(f"Collection: {coll}")

    # Load the collection to get stats
    collection = Collection(name=coll)
    print(collection.num_entities)

# List all collections for coll in utility.list_collections(): print(f"Collection: {coll}") # Load the collection to get stats collection = Collection(name=coll) print(collection.num_entities)

Collection: t2kg_primekg_nodes
2991

In [48]:

%%time

# Assume node_coll_name is defined and collection exists
collection = Collection(node_coll_name)

# Load the collection into memory before query
collection.load()

# Query by expr on node_index
expr = "node_index == 0"
output_fields = ["node_index", "node_id", "node_name", "node_type", "desc", "desc_emb"]

results = collection.query(expr, output_fields=output_fields)

print(results)

%%time # Assume node_coll_name is defined and collection exists collection = Collection(node_coll_name) # Load the collection into memory before query collection.load() # Query by expr on node_index expr = "node_index == 0" output_fields = ["node_index", "node_id", "node_name", "node_type", "desc", "desc_emb"] results = collection.query(expr, output_fields=output_fields) print(results)

data: ['{\'node_index\': 0, \'node_id\': \'SMAD3_(144)\', \'node_name\': \'SMAD3\', \'node_type\': \'gene/protein\', \'desc\': "SMAD3 belongs to gene/protein node. SMAD3 is SMAD family member 3. The SMAD family of proteins are a group of intracellular signal transducer proteins similar to the gene products of the Drosophila gene \'mothers against decapentaplegic\' (Mad) and the C. elegans gene Sma. The SMAD3 protein functions in the transforming growth factor-beta signaling pathway, and transmits signals from the cell surface to the nucleus, regulating gene activity and cell proliferation. This protein forms a complex with other SMAD proteins and binds DNA, functioning both as a transcription factor and tumor suppressor. Mutations in this gene are associated with aneurysms-osteoarthritis syndrome and Loeys-Dietz Syndrome 3. [provided by RefSeq, May 2022].", \'desc_emb\': [-0.036991715, -0.005479036, -0.03023007, -0.012918158, -0.02741491, 0.025599528, -0.024547132, 0.011050156, -0.028256828, -0.0019304885, 0.034308106, 0.0019403548, 0.0078074615, 0.0057158247, -0.0024862853, 0.012885272, 0.021034762, -0.024205104, 0.01377323, 0.020732198, -0.008090293, 0.019416703, -0.02427088, -0.034781683, 0.008794083, 0.034229174, 0.02398147, -0.039701633, -0.015391288, -0.007932434, 0.020653268, 0.0048706196, -0.0089059, 0.021876678, -0.00043699093, 0.011405339, 0.004081323, -0.008070561, 0.012543242, -0.013056286, 0.019035209, 0.0015465285, 0.00945183, -0.014259963, -0.010320056, -0.0123327635, 0.0027411622, -0.032440104, -0.026822938, 0.010024071, -0.0035649908, 0.016035883, -0.031045677, -0.01640422, -0.0077022216, -0.0028102258, 0.016890954, 0.01015562, 0.00947814, -0.010214817, 0.00562374, -0.012477468, -0.029388152, -0.002177144, 0.007261531, 0.011102776, 2.7954264e-05, -0.004939683, -0.011997312, 0.022323947, 0.048673306, 0.03480799, -0.011359298, -0.02307378, -0.0044200625, -0.020074451, -0.03572884, 0.015246584, -0.036465514, -0.0035814345, 0.0044101966, -0.023560511, 0.0060479874, 0.005923015, 0.014154724, 0.0093137035, -0.0008501385, 0.0013985354, -0.0011683238, 0.0073536155, 0.016391065, 0.025652148, -0.0024994402, 0.0052093593, -0.028309448, 0.0027345847, 0.0020538163, 0.032098074, -0.0074062357, -0.015891178, -6.4387306e-05, 0.020942677, -0.015720163, -0.0017874285, -0.041595947, -0.020127071, -0.0057059587, -0.0016838333, 0.026651924, -0.025967866, -0.015141345, 0.04509516, -0.0013080952, -0.018561631, 0.01205651, -0.023968315, 0.029888041, 0.011517157, -0.007787729, 0.0044726823, 0.0065084104, 0.055513877, 0.013865314, -0.03586039, 0.015233429, -0.00037635484, -0.016154276, -0.01012931, -0.025191724, -0.0009619555, 0.014733542, 0.0066794246, 0.015259739, 0.0021738552, -0.012674792, 0.0047357813, -0.00069145695, -0.008063983, -0.01265506, -0.047252573, 0.013944245, 0.016022727, -0.03293999, -0.0070313197, -0.009649155, -0.004873908, 0.03733374, -0.00235638, -0.01645684, -0.0022790947, -0.004498992, -0.009951718, 0.009820169, -0.015036105, 0.014904556, 0.018009124, -0.009616267, 0.019798197, -0.014167879, 0.0046239644, -0.00830735, 0.028967194, -0.004364154, 0.0011321477, 0.010484493, 0.009701774, -0.011477692, -0.0063045085, -0.028914575, -0.0273886, -0.025507445, 0.025691614, -0.0077679968, 0.010392409, -0.023021158, 0.00010698672, -0.0021919433, 0.010905452, -0.018995745, -0.014615146, -0.017232982, -0.009300549, 0.028440997, -0.000105342355, -0.019469323, -0.0055086343, 0.011648706, -0.0064459243, -0.0093729, -0.0046371194, 0.0046239644, 0.011464537, -0.00958338, -0.036044557, -0.61417824, 0.0042095836, 0.009050604, -0.012780032, 0.004814711, 0.04322716, -0.004413485, -0.005663205, 0.0016205251, 0.001856492, -0.02307378, 0.03838614, 0.015088725, -0.022981694, -0.02741491, -0.03338726, -0.012043355, -0.026138881, 0.0475946, -0.016391065, -0.025704768, 0.009307126, -0.00065528083, 0.009497873, -0.004752225, 0.024783922, 0.009576802, -0.030914126, 0.01911414, 0.017193517, -0.017930195, 0.00689977, 0.021205775, 0.015207119, 0.039859492, -0.00042095833, -0.0050942535, 0.04038569, 0.015443908, 0.010931762, -0.004643697, -0.010438452, 0.017824955, 0.0050350563, -0.021679355, 0.031729735, 0.010767325, -0.006084163, -0.003374244, -0.026993953, 0.008623068, 0.007537785, -0.028098969, -0.028519927, 0.0030009726, 0.0072549535, 0.04062248, -0.01653577, 0.036255036, 0.0028316025, -0.008537562, 0.022021383, -0.0190089, 0.007978477, 0.018535322, -0.0147861615, -0.013299652, 0.00752463, -0.0016295691, 0.00684715, 0.012760299, 0.028914575, 0.006649826, 0.011004114, -0.00012281377, 0.05314599, 0.0023810456, 0.044358484, -0.00095537805, 0.011438227, 0.0075180526, -0.0022379856, -0.02254758, -0.016206896, 0.030940436, 0.025218034, -0.023363188, 0.038991265, 0.01879842, -0.0027016974, 0.018061744, -0.0078206165, -0.0034794838, -0.044963613, -0.0035485472, -0.015509684, -0.010274014, 0.030835196, -0.014404668, -0.02382361, -0.006462368, 0.010083267, 0.004278647, -0.017390842, 0.016325291, -0.003979372, -0.039517462, 0.015272894, 0.016825179, -0.017982814, -0.0040221256, -0.00045549005, 0.0037360052, -0.022113467, -0.028283138, -0.011254058, 0.0013960688, -0.010116155, 0.018680027, -0.029782802, 0.010668662, -0.02767801, 0.009978028, 0.010168775, 0.014154724, -0.0008600047, 0.0112014385, -0.010004338, 0.0032032297, -0.010379254, 0.005475747, -0.018403772, -0.005202782, -0.04230631, 0.017127743, -0.003119367, -0.001003887, 0.0045581893, 0.028177898, -0.02358682, -0.0059822127, -0.0039563505, 0.0029828844, -0.021653045, -0.015799092, -0.0028299582, -0.0093137035, 0.013076018, 0.008669111, -0.008649378, -0.018627407, 0.0070444746, 0.005452726, 0.0076890667, 0.020955833, -0.022718595, -0.01187234, -0.004804845, -0.014299428, 0.0025158839, -0.01932462, 0.03020376, -0.02799373, -0.012609017, -0.013444357, -0.03025638, -0.016246362, 0.019429859, -0.00700501, -0.02640198, -0.009787281, -0.021008452, -0.026849248, 0.006991855, 0.0056697824, 0.016548924, 0.0190089, -0.007708799, -0.003699829, 0.0023974893, -0.009182154, -0.029598633, -0.0142204985, -0.010398987, 0.05259348, 0.004301668, 0.021442564, -0.0025356163, 0.0117736785, 0.014878246, -0.031913906, 0.020784818, 0.0010425296, 0.012385383, 0.0017117875, -0.020245465, 0.03520264, -0.0028414687, -0.01929831, 0.016246362, -0.005028479, -0.029282913, 0.0015095302, -0.0019518654, 0.0008912477, -0.03312416, 0.0006778909, -0.013615371, -0.012477468, 0.0006413037, -0.0036208993, 0.005061366, 0.0014363559, 0.014575682, 0.004002393, 0.019942902, -0.00015487896, -0.021771438, -0.0027724053, -0.013496976, -0.008346815, -0.0010556846, 0.018522168, -0.01898259, 0.006965545, 0.018351153, 0.019798197, 0.02799373, 0.0039760834, -0.040096283, -0.010188507, 0.0075312075, 0.014549372, 0.016391065, -0.012891849, -0.0050679436, -0.005192916, -0.01664101, 0.031677116, -0.02252127, -0.013628526, 0.017206673, 0.01887735, -0.0075312075, -0.006064431, -0.017982814, 0.03830721, -0.0028677785, 0.009629422, 0.012885272, -0.015522839, 0.010530536, -0.014509907, -0.006123628, 0.01181972, -0.02637567, 0.00040739228, 0.026691388, 0.055408638, 0.014430977, 0.028704096, -0.01312206, 0.015430753, 0.0075706723, 0.01986397, 0.034492273, 0.0032903813, -0.014062639, 0.034071315, -0.011293523, -0.029177673, 0.01898259, -0.011661861, -0.043937527, 0.007952167, -0.00044932368, 0.013589061, 0.02041648, -0.016733095, 0.02523119, -0.012977356, -0.044595275, 0.02366575, 0.024744457, -0.027704319, 0.0049857255, -0.02398147, -0.006413037, -0.00078518596, 0.029361842, -0.00656103, 0.03304523, 0.018482702, 0.022310792, 0.0030404374, 0.0053573526, 0.012780032, 0.0049331053, 0.021192621, 0.0046502743, 0.0024566865, 0.0075904047, -0.0014766429, -0.016233206, 0.038123038, 0.002200165, 0.0026605881, -0.0011617463, -0.022218706, -0.008695421, -0.0024484647, -0.036623377, -0.009682042, -0.014641456, -0.0019173336, -0.0010153976, 0.0044693938, 0.0076167146, 0.015522839, 0.0044496614, -0.018048588, -0.04283251, -0.00042465815, 0.014615146, 0.0043444214, 0.046068627, -0.0044332175, -0.0004195195, 0.00017471415, 0.0042589144, -0.02744122, 0.0075838272, 0.0190089, -0.01018193, -0.023455271, -0.0011642129, 0.015430753, -0.008899323, -0.0045351684, 0.022744905, 0.0032755819, -0.010201662, 0.014799316, -0.005189627, 0.0014240231, -0.0068931924, 0.019206224, 0.0356236, 0.03491323, 0.0005286645, -0.019232534, 0.028783025, 0.024073554, -0.022560736, -0.003923463, 0.021311015, 0.000589095, 0.018114364, -0.013247033, -0.008583603, 0.015733318, 0.02650722, -0.021718819, 0.023652596, 0.023008004, 0.016325291, 0.024770766, 0.0029713737, -0.02512595, -0.00689977, -0.017127743, -0.0139574, 0.006172959, -0.009846479, 0.00016299803, -0.041806426, 0.0020965699, -0.022836989, 0.04541088, -0.041332845, -0.0034202863, -0.017969659, 0.002338292, -0.026691388, -0.010602888, -0.011832875, 0.0030782577, -0.005380374, -0.00035189485, -0.014259963, 0.012872117, 0.005867107, -0.019969212, -0.0106423525, 0.009682042, -0.00749832, -0.03372929, -0.0073536155, 0.009109802, 0.026651924, 0.020192845, -0.008379702, -0.0069852774, 0.039280675, 0.00040903664, -0.033755597, 0.0043049566, -0.019956056, -0.0013730477, 0.015391288, 0.008419167, -0.0027230743, -0.013944245, -0.0007716199, 0.011602664, 0.007379926, -0.008031096, -0.020771664, -0.008583603, -0.021666199, 0.031861287, 0.0021672777, 0.0042589144, -0.0072944183, 0.001271919, -0.030466858, -0.028204208, -0.013003666, -0.0064064595, -0.004265492, 0.013450934, 2.1107011e-05, -0.011267213, -0.012405116, -0.010149042, 0.0113921845, 0.011096198, -0.017561857, 0.017114587, 0.017943349, 0.013220723, 0.0056599164, -0.0004850887, -0.006038121, -0.012786609, -0.009945141, 0.0029664407, 0.010662085, -0.019469323, 0.017824955, -0.0024287323, -0.01265506, -0.016009573, 0.008175801, 0.020455943, 0.039201744, 0.008701998, -0.012161749, -0.009484718, 0.00028365356, -0.016890954, -0.0054297047, -0.0078206165, -0.013891624, -0.0019387105, 0.01023455, 0.011642129, -0.017588167, -0.0038083573, -0.021876678, -0.028414687, 0.0039596395, 0.027151812, 0.0119578475, -0.010241127, -0.011688171, -0.00269512, -0.010918607, -0.010859409, -0.046963163, -0.0057816, 0.0005315421, 0.028625166, 0.034413345, 0.02158727, 0.0006108829, 0.02270544, 0.023350032, 0.010958072, 0.0052685565, -0.0015021306, 0.0053869514, -0.01653577, 0.014549372, 0.0045450344, 0.0067419107, 0.023047468, -0.005636895, 0.023336878, 0.017009348, -0.0026211233, -0.008484942, 0.044148006, -0.03825459, -0.010379254, 0.007379926, 0.0076298695, 0.0089059, -0.025941556, -0.009958296, 0.008169223, 0.028493617, -0.009122957, 0.0043477104, 0.014115259, -0.022258172, 0.041359156, -0.0069852774, 0.015641233, 0.000544286, -0.0042194496, 0.0036307655, -0.016759405, 0.004489126, -0.009215041, 0.02028493, -0.0015341957, 0.015391288, -0.0045088585, 0.016746249, -0.024047244, -0.032308552, -0.011655284, -0.011497424, -0.012720834, -0.041517016, -0.02377099, -0.03864924, 0.005314599, 0.024599751, -0.009760971, 0.021389944, 0.008550717, -0.006659692, 0.029309222, 0.00593617, 0.034097627, -0.01656208, 0.026638769, 0.019706111, 0.014628301, -0.017127743, -0.0067550656, -0.0023744681, -0.00026433225, 0.0214031, -0.0028957329, -0.012582707, 0.006100607, -0.004189851, -0.007149714, -0.015614923, -0.010576578, 0.018758956, -0.010372677, -0.003673519, 0.0030108388, 0.013812695, 0.00828104, 0.00089453644, -0.0038379559, -0.00044726822, -0.012431426, -0.012398538, -0.016127966, 0.024626061, 0.009550492, 0.021074226, -0.0029368422, -0.007833771, 0.0102674365, -0.015838558, 0.011089621, -0.002168922, 0.012569552, 0.041359156, 0.028572546, -0.013358849, 0.0066958684, 0.017614476, -0.00035847232, -0.010372677, -0.005278423, 0.034439653, -0.009280816, -0.0151676545, 0.010523958, -0.0100635355, -0.0042062947, 0.0120302, -0.0025504155, -0.016825179, -0.03559729, 0.006024966, 0.021863524, -0.0025471267, -0.009931985, 0.019416703, -0.029546013, -0.02392885, -0.03262427, -0.013016821, 0.012115707, -0.061038956, -0.0049100844, -0.0039201747, 0.017614476, 0.019535098, 0.029361842, -0.025375893, 0.020100761, 0.03304523, -0.020666422, 0.0043345555, -0.0025833028, -0.0055316556, -0.0146019915, 0.020061295, -0.01871949, -0.04270096, 0.008609913, 0.011984157, -0.015404443, -0.0040352806, -0.0042918017, -0.003020705, -0.005548099, -0.03333464, 0.03567622, -0.021600425, 0.009168999, -0.032492723, -0.010188507, 0.017509237, -0.019403549, 0.0093729, -0.005840797, 0.017298756, -0.0031390993, 0.012516933, -0.026717698, 0.0087546185, -0.0004637119, 0.002780627, -0.01401002, 0.018259067, -0.000803274, 0.022823835, -0.0031621205, 0.009905675, -0.009215041, 0.010662085, 0.013141793, -0.014536217, -0.015930643, 0.019285154, -0.029598633, 0.0016698561, -0.024257723, -0.004949549, 0.0009890876, -0.031729735, -0.03023007, 0.0012488979, -0.039385915, 0.006442636, 0.016443685, 0.002861201, -0.012556397, 0.002039017, -0.03299261, -0.02153465, -0.017903885, -0.0006992677, -0.0028628455, 0.035228953, -0.014588837, 0.007176024, -0.0066399598, -0.0001528235, -0.019337773, -0.008116603, 0.0059427475, -0.021389944, 0.0030108388, 0.006630094, -0.021981917, -0.015904333, 0.024599751, 0.018219603, -0.008287617, 0.01932462, -0.052567173, 0.019614028, -0.024231413, 0.006149938, -0.025494289, 0.008715153, 0.021284705, -0.027099192, -0.0050712326, 0.002155767, 0.009024294, -0.014246808, 0.010912029, -0.00042794688, 0.00661365, -0.004015548, 0.024152484, -0.008320505, 0.0020110628, 0.0190089, 0.025757387, 0.028388377, 0.006712312, 0.025757387, 0.021140002, -0.0013286497, -0.004587788, -0.016496304, -0.0045187245, -0.010662085, -0.0091361115, 0.0043477104, -0.025375893, -0.027257051, -0.017061967, -0.032334864, -0.005137007, -0.012661637, -0.00079011905, -0.025270654, 0.014694076, -0.00825473, -0.02033755, -0.0017348087, 0.000118291755, -0.022508116, -0.017088277, -0.005969058, 0.009991183, -0.012010467, -0.023165863, 0.004324689, -0.008123181, 0.002530683, 0.0006540476, -0.007945589, -0.015062415, 0.0025191726, 0.20111284, -0.0038642657, -0.0146019915, 0.014299428, 0.02509964, 0.036176108, 0.038780786, -0.0034400187, -0.004949549, 0.0016608122, 0.011602664, 0.011885495, 0.006965545, 0.0011617463, 0.0008213621, -0.016785715, -0.026415136, 0.0013154948, -0.030756267, -0.01914045, 0.002538905, -0.0053113103, -0.024336653, -0.0042490484, 0.013589061, 0.009432098, -0.04072772, -0.0032525607, 0.025507445, 0.015338669, -0.02523119, -0.00052003155, 0.018732646, -0.003926752, 0.004101055, 0.001003887, -0.009412365, -0.020758508, 0.019732421, 0.02358682, -0.008583603, 0.004939683, -0.014312583, -0.010037226, 0.011227748, 0.008951942, -0.0029351977, 0.008465209, -0.020140225, 0.002690187, -0.017325066, 0.010214817, -0.0044101966, 0.014444132, -0.008215265, 0.00016022315, 0.014115259, 0.016838335, 0.013286497, 0.010379254, -0.017995968, -0.011332988, -0.03309785, 0.014930866, -0.0072681084, 0.013233878, -0.006314375, 0.026033642, 0.022981694, -0.0452004, -0.0035584134, -0.01265506, -0.010938339, -0.011004114, -0.024481358, 0.0065511637, 0.05033083, 0.002900666, 0.019127294, 0.018232757, -0.008550717, -0.022100313, -0.028809335, 0.016575234, 0.004873908, -0.011076466, -0.0024731301, -0.007721954, 0.00066021393, -0.025033865, 0.0095110275, -0.0132799195, -0.0009759327, 0.0061466494, 0.006186114, 0.007123404, 0.011175129, 0.0015958596, 0.008169223, -0.006084163, -0.00596248, 0.06972122, 0.004189851, 0.015825402, 0.005248824, 0.017101433, -0.004804845, 0.012701102, 0.008070561, -0.03357143, 0.004239182, -0.026309896, -0.011431649, -0.00013802419, 0.013233878, 0.011543467, 0.029493393, -0.032071766, 0.015246584, -0.009011139, 0.004301668, -0.0102608595, -0.0055579655, 0.023573667, 0.0034169976, 0.015457063, -0.0065544527, -0.017798645, 0.00687346, -0.025770543, -0.0008830259, 0.00117819, 0.0004480904, 0.01382585, 0.016627854, -0.020705888, 0.008432322, 0.018548477, -0.003400554, 0.026257277, -0.022455497, 0.0023794011, 0.021705665, 0.0068076854, -0.0044101966, -0.007024742, 0.020995297, 0.02377099, -0.013733765, -0.009497873, -0.024441892, -0.00833366, 0.010024071, -0.011471114, 0.0112014385, -0.019311463, -0.004890352, -0.012911581, 0.010688395, 0.004337844, -0.024639217, 0.011530312, 0.008221842, -0.017982814, -0.012365651, -0.0012414982, -0.16417375, 0.006186114, 0.0166147, -0.034939542, 0.005564543, 0.024941782, 0.03343988, 0.0060151, 0.0089059, 0.0005113986, 0.03754422, 0.022218706, -0.023915695, -0.0020044853, -0.0061795367, 0.024284033, 0.0151676545, 0.045963388, -0.0050975424, 0.023021158, 0.002637567, -0.015825402, 0.011425072, -0.011497424, 0.008728308, 0.016259516, 0.013286497, -0.036333967, -0.031124607, 0.01135272, -0.0019469323, 0.0051304298, 0.023284258, 0.0006281488, 0.007176024, 0.01020824, 0.0073930807, 9.645249e-05, -0.016022727, 0.024560288, 0.0023826899, 0.030440548, -0.0050087464, 0.009767549, -0.016417375, 0.03017745, 0.005426416, -0.0010819945, 0.027204432, -0.0073601934, 0.02523119, -0.013760075, 0.0027526729, 0.037281122, 0.039596394, 0.019508788, -0.0023662462, 0.021126846, -0.0069852774, -0.024323499, -0.020377014, -0.017456615, 0.031966526, -0.01012931, 0.0056566275, -0.020034986, -0.017206673, 0.032334864, -0.02799373, 0.016785715, -0.029703872, 0.0049692816, 0.0030305712, -0.010958072, 0.004110921, 0.0113001, -0.023389498, 0.014352048, -0.00049906585, 0.0010244416, -0.013418047, 0.00440033, -0.031598184, -0.012747144, 0.002713208, 0.0071957563, 0.0117736785, 0.007051052, 0.004485837, -0.0144178225, 0.016706785, -0.007202334, 0.012477468, -0.016075347, 0.026559839, 0.011155396, -0.018680027, -0.018180138, -0.023257948, -0.006238734, 0.019127294, -0.0123261865, -0.009076914, 0.016601544, 0.011596086, 0.016825179, -0.022350257, 0.014207344, 0.040701408, -0.0136548355, 0.0042194496, -0.017561857, 0.013424624, 0.005663205, -0.0013228945, 0.026086261, -0.006001945, -0.010043803, 0.014812471, -0.012497201, 0.06745858, 0.025586374, 0.01986397, 0.0151545, -0.017219827, -0.02310009, -0.10839677, -0.01892997, 0.008438899, 0.03309785, -0.049409986, -0.004137231, -0.0085638715, 0.033860836, -0.014865091, 0.029730182, -0.028993504, -0.029203983, 0.0018532033, -0.008991407, 0.0048936405, -0.010135887, 0.01760132, -0.007248376, -0.017482925, 0.03830721, -0.011734214, -0.016035883, 0.016114812, -0.007136559, 0.020587493, 0.016851489, -0.03333464, 0.0062354454, -3.029235e-05, 0.0073141507, -0.0031703424, -0.0300459, 0.04848914, 0.01650946, -0.007721954, -0.022416031, 0.009655732, -0.02145572, 0.04788401, 0.0008805593, 0.029177673, 0.0002542605, -0.011865763, -0.036886476, -0.0078206165, -0.014391513, -0.039649013, 0.031492945, 0.022639666, -0.016009573, -0.033992387, -0.0013171391, -0.027730629, -0.005038345, 0.028467307, -0.009741239, 0.031545565, 0.0047226264, -0.01124748, 0.007149714, -0.009997761, 0.012457736, 0.013247033, -0.0071562915, 0.005334331, -0.029256603, -0.014352048, -0.056829374, 0.054303624, -0.01869318, 0.008767773, -0.006965545, -0.014575682, -0.01320099, -0.027072882, 0.027730629, -0.011411917, -0.025560064, 0.014457287, -0.038780786, -0.004716049, -0.01502295, -0.0041701184, -0.017338222, 0.018864196, 0.019403549, 0.008846703, -0.013878469, 0.023876231, -0.037412673, 0.012457736, 0.03025638, -0.001290007, 0.00040410354, -0.01393109, 0.012017045, -0.007873236, -0.0064064595, 0.0166147, 0.033308327, -0.030808887, -0.033360947, -0.06577474, 0.013339117, 0.0007350327, 0.0065676076, 0.027309671, -0.026888713, 0.021705665, -0.008524407, 0.0048245774, -0.010885719, 0.0020094183, -0.004341133, -0.0155359935, -0.0038576883, -0.009649155, -0.014049484, 0.03849138, -0.009333435, 0.030703647, 0.03280844, 0.0035945894, -0.012115707, 0.01382585, 0.039649013, -0.0029565745, -0.0069063473, -0.00010914496, 0.007892969, 0.0015687275, 0.023521047, -0.014904556, -0.037096955, -0.011457959, 0.005627029, -0.018153828, -0.039649013, -0.026783474, -0.003400554, 0.019824507, 0.028230518, -0.0031341664, -0.04764722, 0.024836542, -0.012490623, 0.018364308, 0.003288737, -0.024205104, -0.009708351, -0.0057026697, -0.009793859, 0.031545565, 0.008846703, 0.009392633, -0.025560064, -0.022376567, -0.012964201, 0.015878022, -6.8189904e-05, -0.022810679, -0.003824801, 0.02733598, 0.025967866, 0.010570001, 0.012102552, 0.0011995669, -0.008484942, -0.024928626, 0.02025862, 0.0050515, -0.015496528, -0.017193517, -0.005913149, 0.0079982085, 0.011359298, 0.0038379559, 0.0078206165, -0.028967194, -0.024481358, -0.024573442, 0.028256828, 0.029230293, 0.0017035657, -0.008471787, 0.028914575, 0.014838781, -0.0065347203, -0.032229625, 0.014299428, 0.016088502, 0.006429481, -0.014233653, 0.041464396, 0.007912701, -0.0076035596, 0.009017717, 0.023402652, 0.017509237, 0.016917264, 0.01312206, 0.036255036, 0.001963376, -0.003660364, -0.015601768, 0.001428134, -0.022245016, 0.009201886, -0.022008227, -0.04814711, -0.0056072967, -0.004390464, 0.012352496, -0.014667766, 0.0041569634, 0.0071299816, -0.003877421, 0.013214145, 0.024797076, -0.018811576, -0.011063311, 0.025691614, 0.012668215, 0.0027362292, -0.020245465, -0.00051386515, 0.031834975, -0.000110378234, 0.033913456, -0.04327978, 0.020995297, -0.032150693, -0.025849473, 0.00039074305, -0.034281794, 0.0001238415, 0.009563647, -0.008616491, -0.0067188893, 0.011069888, -0.001682189, 0.083086655, 0.0070181647, -0.008412589, -0.022692285, -0.011115931, 0.039070196, 0.007860081, 0.007886391, 0.00033257352, -0.02653353, -0.007425968, -0.03491323, -0.0069852774, -0.016049037, -0.019272, 0.041517016, -0.015956953, 0.023942005, 0.0043214005, 0.0005685404, 0.009215041, -0.016733095, -0.020021832, 0.0004320578, -0.036228728, 0.0004008148, 0.023376342, 0.005814487, 0.011319833, -0.037070643, -0.012253834, 0.0035814345, -0.046831615, -0.009800436, 0.0009290682, -0.018640561, 0.01929831, -0.024468202, 0.0030009726, 0.026691388, 0.010221395, -0.0022412743, -0.0020982143, -0.02759908, 0.01671994, -0.0028496906, -0.0021804327, -0.004265492, -0.006761643]}']
CPU times: user 11.4 ms, sys: 2.55 ms, total: 13.9 ms
Wall time: 1.49 s

In [49]:

%%time

# Assume node_coll_name is defined and collection exists
collection = Collection(node_coll_name)

# Load the collection into memory before query
collection.load()

# Query by expr on node_index
expr = "node_index in [0, 1]"
output_fields = ["node_index", "node_id", "node_name", "node_type", "desc", "desc_emb"]

results = collection.query(expr, output_fields=output_fields)

print(results)

%%time # Assume node_coll_name is defined and collection exists collection = Collection(node_coll_name) # Load the collection into memory before query collection.load() # Query by expr on node_index expr = "node_index in [0, 1]" output_fields = ["node_index", "node_id", "node_name", "node_type", "desc", "desc_emb"] results = collection.query(expr, output_fields=output_fields) print(results)

data: ['{\'node_name\': \'SMAD3\', \'node_type\': \'gene/protein\', \'desc\': "SMAD3 belongs to gene/protein node. SMAD3 is SMAD family member 3. The SMAD family of proteins are a group of intracellular signal transducer proteins similar to the gene products of the Drosophila gene \'mothers against decapentaplegic\' (Mad) and the C. elegans gene Sma. The SMAD3 protein functions in the transforming growth factor-beta signaling pathway, and transmits signals from the cell surface to the nucleus, regulating gene activity and cell proliferation. This protein forms a complex with other SMAD proteins and binds DNA, functioning both as a transcription factor and tumor suppressor. Mutations in this gene are associated with aneurysms-osteoarthritis syndrome and Loeys-Dietz Syndrome 3. [provided by RefSeq, May 2022].", \'desc_emb\': [-0.036991715, -0.005479036, -0.03023007, -0.012918158, -0.02741491, 0.025599528, -0.024547132, 0.011050156, -0.028256828, -0.0019304885, 0.034308106, 0.0019403548, 0.0078074615, 0.0057158247, -0.0024862853, 0.012885272, 0.021034762, -0.024205104, 0.01377323, 0.020732198, -0.008090293, 0.019416703, -0.02427088, -0.034781683, 0.008794083, 0.034229174, 0.02398147, -0.039701633, -0.015391288, -0.007932434, 0.020653268, 0.0048706196, -0.0089059, 0.021876678, -0.00043699093, 0.011405339, 0.004081323, -0.008070561, 0.012543242, -0.013056286, 0.019035209, 0.0015465285, 0.00945183, -0.014259963, -0.010320056, -0.0123327635, 0.0027411622, -0.032440104, -0.026822938, 0.010024071, -0.0035649908, 0.016035883, -0.031045677, -0.01640422, -0.0077022216, -0.0028102258, 0.016890954, 0.01015562, 0.00947814, -0.010214817, 0.00562374, -0.012477468, -0.029388152, -0.002177144, 0.007261531, 0.011102776, 2.7954264e-05, -0.004939683, -0.011997312, 0.022323947, 0.048673306, 0.03480799, -0.011359298, -0.02307378, -0.0044200625, -0.020074451, -0.03572884, 0.015246584, -0.036465514, -0.0035814345, 0.0044101966, -0.023560511, 0.0060479874, 0.005923015, 0.014154724, 0.0093137035, -0.0008501385, 0.0013985354, -0.0011683238, 0.0073536155, 0.016391065, 0.025652148, -0.0024994402, 0.0052093593, -0.028309448, 0.0027345847, 0.0020538163, 0.032098074, -0.0074062357, -0.015891178, -6.4387306e-05, 0.020942677, -0.015720163, -0.0017874285, -0.041595947, -0.020127071, -0.0057059587, -0.0016838333, 0.026651924, -0.025967866, -0.015141345, 0.04509516, -0.0013080952, -0.018561631, 0.01205651, -0.023968315, 0.029888041, 0.011517157, -0.007787729, 0.0044726823, 0.0065084104, 0.055513877, 0.013865314, -0.03586039, 0.015233429, -0.00037635484, -0.016154276, -0.01012931, -0.025191724, -0.0009619555, 0.014733542, 0.0066794246, 0.015259739, 0.0021738552, -0.012674792, 0.0047357813, -0.00069145695, -0.008063983, -0.01265506, -0.047252573, 0.013944245, 0.016022727, -0.03293999, -0.0070313197, -0.009649155, -0.004873908, 0.03733374, -0.00235638, -0.01645684, -0.0022790947, -0.004498992, -0.009951718, 0.009820169, -0.015036105, 0.014904556, 0.018009124, -0.009616267, 0.019798197, -0.014167879, 0.0046239644, -0.00830735, 0.028967194, -0.004364154, 0.0011321477, 0.010484493, 0.009701774, -0.011477692, -0.0063045085, -0.028914575, -0.0273886, -0.025507445, 0.025691614, -0.0077679968, 0.010392409, -0.023021158, 0.00010698672, -0.0021919433, 0.010905452, -0.018995745, -0.014615146, -0.017232982, -0.009300549, 0.028440997, -0.000105342355, -0.019469323, -0.0055086343, 0.011648706, -0.0064459243, -0.0093729, -0.0046371194, 0.0046239644, 0.011464537, -0.00958338, -0.036044557, -0.61417824, 0.0042095836, 0.009050604, -0.012780032, 0.004814711, 0.04322716, -0.004413485, -0.005663205, 0.0016205251, 0.001856492, -0.02307378, 0.03838614, 0.015088725, -0.022981694, -0.02741491, -0.03338726, -0.012043355, -0.026138881, 0.0475946, -0.016391065, -0.025704768, 0.009307126, -0.00065528083, 0.009497873, -0.004752225, 0.024783922, 0.009576802, -0.030914126, 0.01911414, 0.017193517, -0.017930195, 0.00689977, 0.021205775, 0.015207119, 0.039859492, -0.00042095833, -0.0050942535, 0.04038569, 0.015443908, 0.010931762, -0.004643697, -0.010438452, 0.017824955, 0.0050350563, -0.021679355, 0.031729735, 0.010767325, -0.006084163, -0.003374244, -0.026993953, 0.008623068, 0.007537785, -0.028098969, -0.028519927, 0.0030009726, 0.0072549535, 0.04062248, -0.01653577, 0.036255036, 0.0028316025, -0.008537562, 0.022021383, -0.0190089, 0.007978477, 0.018535322, -0.0147861615, -0.013299652, 0.00752463, -0.0016295691, 0.00684715, 0.012760299, 0.028914575, 0.006649826, 0.011004114, -0.00012281377, 0.05314599, 0.0023810456, 0.044358484, -0.00095537805, 0.011438227, 0.0075180526, -0.0022379856, -0.02254758, -0.016206896, 0.030940436, 0.025218034, -0.023363188, 0.038991265, 0.01879842, -0.0027016974, 0.018061744, -0.0078206165, -0.0034794838, -0.044963613, -0.0035485472, -0.015509684, -0.010274014, 0.030835196, -0.014404668, -0.02382361, -0.006462368, 0.010083267, 0.004278647, -0.017390842, 0.016325291, -0.003979372, -0.039517462, 0.015272894, 0.016825179, -0.017982814, -0.0040221256, -0.00045549005, 0.0037360052, -0.022113467, -0.028283138, -0.011254058, 0.0013960688, -0.010116155, 0.018680027, -0.029782802, 0.010668662, -0.02767801, 0.009978028, 0.010168775, 0.014154724, -0.0008600047, 0.0112014385, -0.010004338, 0.0032032297, -0.010379254, 0.005475747, -0.018403772, -0.005202782, -0.04230631, 0.017127743, -0.003119367, -0.001003887, 0.0045581893, 0.028177898, -0.02358682, -0.0059822127, -0.0039563505, 0.0029828844, -0.021653045, -0.015799092, -0.0028299582, -0.0093137035, 0.013076018, 0.008669111, -0.008649378, -0.018627407, 0.0070444746, 0.005452726, 0.0076890667, 0.020955833, -0.022718595, -0.01187234, -0.004804845, -0.014299428, 0.0025158839, -0.01932462, 0.03020376, -0.02799373, -0.012609017, -0.013444357, -0.03025638, -0.016246362, 0.019429859, -0.00700501, -0.02640198, -0.009787281, -0.021008452, -0.026849248, 0.006991855, 0.0056697824, 0.016548924, 0.0190089, -0.007708799, -0.003699829, 0.0023974893, -0.009182154, -0.029598633, -0.0142204985, -0.010398987, 0.05259348, 0.004301668, 0.021442564, -0.0025356163, 0.0117736785, 0.014878246, -0.031913906, 0.020784818, 0.0010425296, 0.012385383, 0.0017117875, -0.020245465, 0.03520264, -0.0028414687, -0.01929831, 0.016246362, -0.005028479, -0.029282913, 0.0015095302, -0.0019518654, 0.0008912477, -0.03312416, 0.0006778909, -0.013615371, -0.012477468, 0.0006413037, -0.0036208993, 0.005061366, 0.0014363559, 0.014575682, 0.004002393, 0.019942902, -0.00015487896, -0.021771438, -0.0027724053, -0.013496976, -0.008346815, -0.0010556846, 0.018522168, -0.01898259, 0.006965545, 0.018351153, 0.019798197, 0.02799373, 0.0039760834, -0.040096283, -0.010188507, 0.0075312075, 0.014549372, 0.016391065, -0.012891849, -0.0050679436, -0.005192916, -0.01664101, 0.031677116, -0.02252127, -0.013628526, 0.017206673, 0.01887735, -0.0075312075, -0.006064431, -0.017982814, 0.03830721, -0.0028677785, 0.009629422, 0.012885272, -0.015522839, 0.010530536, -0.014509907, -0.006123628, 0.01181972, -0.02637567, 0.00040739228, 0.026691388, 0.055408638, 0.014430977, 0.028704096, -0.01312206, 0.015430753, 0.0075706723, 0.01986397, 0.034492273, 0.0032903813, -0.014062639, 0.034071315, -0.011293523, -0.029177673, 0.01898259, -0.011661861, -0.043937527, 0.007952167, -0.00044932368, 0.013589061, 0.02041648, -0.016733095, 0.02523119, -0.012977356, -0.044595275, 0.02366575, 0.024744457, -0.027704319, 0.0049857255, -0.02398147, -0.006413037, -0.00078518596, 0.029361842, -0.00656103, 0.03304523, 0.018482702, 0.022310792, 0.0030404374, 0.0053573526, 0.012780032, 0.0049331053, 0.021192621, 0.0046502743, 0.0024566865, 0.0075904047, -0.0014766429, -0.016233206, 0.038123038, 0.002200165, 0.0026605881, -0.0011617463, -0.022218706, -0.008695421, -0.0024484647, -0.036623377, -0.009682042, -0.014641456, -0.0019173336, -0.0010153976, 0.0044693938, 0.0076167146, 0.015522839, 0.0044496614, -0.018048588, -0.04283251, -0.00042465815, 0.014615146, 0.0043444214, 0.046068627, -0.0044332175, -0.0004195195, 0.00017471415, 0.0042589144, -0.02744122, 0.0075838272, 0.0190089, -0.01018193, -0.023455271, -0.0011642129, 0.015430753, -0.008899323, -0.0045351684, 0.022744905, 0.0032755819, -0.010201662, 0.014799316, -0.005189627, 0.0014240231, -0.0068931924, 0.019206224, 0.0356236, 0.03491323, 0.0005286645, -0.019232534, 0.028783025, 0.024073554, -0.022560736, -0.003923463, 0.021311015, 0.000589095, 0.018114364, -0.013247033, -0.008583603, 0.015733318, 0.02650722, -0.021718819, 0.023652596, 0.023008004, 0.016325291, 0.024770766, 0.0029713737, -0.02512595, -0.00689977, -0.017127743, -0.0139574, 0.006172959, -0.009846479, 0.00016299803, -0.041806426, 0.0020965699, -0.022836989, 0.04541088, -0.041332845, -0.0034202863, -0.017969659, 0.002338292, -0.026691388, -0.010602888, -0.011832875, 0.0030782577, -0.005380374, -0.00035189485, -0.014259963, 0.012872117, 0.005867107, -0.019969212, -0.0106423525, 0.009682042, -0.00749832, -0.03372929, -0.0073536155, 0.009109802, 0.026651924, 0.020192845, -0.008379702, -0.0069852774, 0.039280675, 0.00040903664, -0.033755597, 0.0043049566, -0.019956056, -0.0013730477, 0.015391288, 0.008419167, -0.0027230743, -0.013944245, -0.0007716199, 0.011602664, 0.007379926, -0.008031096, -0.020771664, -0.008583603, -0.021666199, 0.031861287, 0.0021672777, 0.0042589144, -0.0072944183, 0.001271919, -0.030466858, -0.028204208, -0.013003666, -0.0064064595, -0.004265492, 0.013450934, 2.1107011e-05, -0.011267213, -0.012405116, -0.010149042, 0.0113921845, 0.011096198, -0.017561857, 0.017114587, 0.017943349, 0.013220723, 0.0056599164, -0.0004850887, -0.006038121, -0.012786609, -0.009945141, 0.0029664407, 0.010662085, -0.019469323, 0.017824955, -0.0024287323, -0.01265506, -0.016009573, 0.008175801, 0.020455943, 0.039201744, 0.008701998, -0.012161749, -0.009484718, 0.00028365356, -0.016890954, -0.0054297047, -0.0078206165, -0.013891624, -0.0019387105, 0.01023455, 0.011642129, -0.017588167, -0.0038083573, -0.021876678, -0.028414687, 0.0039596395, 0.027151812, 0.0119578475, -0.010241127, -0.011688171, -0.00269512, -0.010918607, -0.010859409, -0.046963163, -0.0057816, 0.0005315421, 0.028625166, 0.034413345, 0.02158727, 0.0006108829, 0.02270544, 0.023350032, 0.010958072, 0.0052685565, -0.0015021306, 0.0053869514, -0.01653577, 0.014549372, 0.0045450344, 0.0067419107, 0.023047468, -0.005636895, 0.023336878, 0.017009348, -0.0026211233, -0.008484942, 0.044148006, -0.03825459, -0.010379254, 0.007379926, 0.0076298695, 0.0089059, -0.025941556, -0.009958296, 0.008169223, 0.028493617, -0.009122957, 0.0043477104, 0.014115259, -0.022258172, 0.041359156, -0.0069852774, 0.015641233, 0.000544286, -0.0042194496, 0.0036307655, -0.016759405, 0.004489126, -0.009215041, 0.02028493, -0.0015341957, 0.015391288, -0.0045088585, 0.016746249, -0.024047244, -0.032308552, -0.011655284, -0.011497424, -0.012720834, -0.041517016, -0.02377099, -0.03864924, 0.005314599, 0.024599751, -0.009760971, 0.021389944, 0.008550717, -0.006659692, 0.029309222, 0.00593617, 0.034097627, -0.01656208, 0.026638769, 0.019706111, 0.014628301, -0.017127743, -0.0067550656, -0.0023744681, -0.00026433225, 0.0214031, -0.0028957329, -0.012582707, 0.006100607, -0.004189851, -0.007149714, -0.015614923, -0.010576578, 0.018758956, -0.010372677, -0.003673519, 0.0030108388, 0.013812695, 0.00828104, 0.00089453644, -0.0038379559, -0.00044726822, -0.012431426, -0.012398538, -0.016127966, 0.024626061, 0.009550492, 0.021074226, -0.0029368422, -0.007833771, 0.0102674365, -0.015838558, 0.011089621, -0.002168922, 0.012569552, 0.041359156, 0.028572546, -0.013358849, 0.0066958684, 0.017614476, -0.00035847232, -0.010372677, -0.005278423, 0.034439653, -0.009280816, -0.0151676545, 0.010523958, -0.0100635355, -0.0042062947, 0.0120302, -0.0025504155, -0.016825179, -0.03559729, 0.006024966, 0.021863524, -0.0025471267, -0.009931985, 0.019416703, -0.029546013, -0.02392885, -0.03262427, -0.013016821, 0.012115707, -0.061038956, -0.0049100844, -0.0039201747, 0.017614476, 0.019535098, 0.029361842, -0.025375893, 0.020100761, 0.03304523, -0.020666422, 0.0043345555, -0.0025833028, -0.0055316556, -0.0146019915, 0.020061295, -0.01871949, -0.04270096, 0.008609913, 0.011984157, -0.015404443, -0.0040352806, -0.0042918017, -0.003020705, -0.005548099, -0.03333464, 0.03567622, -0.021600425, 0.009168999, -0.032492723, -0.010188507, 0.017509237, -0.019403549, 0.0093729, -0.005840797, 0.017298756, -0.0031390993, 0.012516933, -0.026717698, 0.0087546185, -0.0004637119, 0.002780627, -0.01401002, 0.018259067, -0.000803274, 0.022823835, -0.0031621205, 0.009905675, -0.009215041, 0.010662085, 0.013141793, -0.014536217, -0.015930643, 0.019285154, -0.029598633, 0.0016698561, -0.024257723, -0.004949549, 0.0009890876, -0.031729735, -0.03023007, 0.0012488979, -0.039385915, 0.006442636, 0.016443685, 0.002861201, -0.012556397, 0.002039017, -0.03299261, -0.02153465, -0.017903885, -0.0006992677, -0.0028628455, 0.035228953, -0.014588837, 0.007176024, -0.0066399598, -0.0001528235, -0.019337773, -0.008116603, 0.0059427475, -0.021389944, 0.0030108388, 0.006630094, -0.021981917, -0.015904333, 0.024599751, 0.018219603, -0.008287617, 0.01932462, -0.052567173, 0.019614028, -0.024231413, 0.006149938, -0.025494289, 0.008715153, 0.021284705, -0.027099192, -0.0050712326, 0.002155767, 0.009024294, -0.014246808, 0.010912029, -0.00042794688, 0.00661365, -0.004015548, 0.024152484, -0.008320505, 0.0020110628, 0.0190089, 0.025757387, 0.028388377, 0.006712312, 0.025757387, 0.021140002, -0.0013286497, -0.004587788, -0.016496304, -0.0045187245, -0.010662085, -0.0091361115, 0.0043477104, -0.025375893, -0.027257051, -0.017061967, -0.032334864, -0.005137007, -0.012661637, -0.00079011905, -0.025270654, 0.014694076, -0.00825473, -0.02033755, -0.0017348087, 0.000118291755, -0.022508116, -0.017088277, -0.005969058, 0.009991183, -0.012010467, -0.023165863, 0.004324689, -0.008123181, 0.002530683, 0.0006540476, -0.007945589, -0.015062415, 0.0025191726, 0.20111284, -0.0038642657, -0.0146019915, 0.014299428, 0.02509964, 0.036176108, 0.038780786, -0.0034400187, -0.004949549, 0.0016608122, 0.011602664, 0.011885495, 0.006965545, 0.0011617463, 0.0008213621, -0.016785715, -0.026415136, 0.0013154948, -0.030756267, -0.01914045, 0.002538905, -0.0053113103, -0.024336653, -0.0042490484, 0.013589061, 0.009432098, -0.04072772, -0.0032525607, 0.025507445, 0.015338669, -0.02523119, -0.00052003155, 0.018732646, -0.003926752, 0.004101055, 0.001003887, -0.009412365, -0.020758508, 0.019732421, 0.02358682, -0.008583603, 0.004939683, -0.014312583, -0.010037226, 0.011227748, 0.008951942, -0.0029351977, 0.008465209, -0.020140225, 0.002690187, -0.017325066, 0.010214817, -0.0044101966, 0.014444132, -0.008215265, 0.00016022315, 0.014115259, 0.016838335, 0.013286497, 0.010379254, -0.017995968, -0.011332988, -0.03309785, 0.014930866, -0.0072681084, 0.013233878, -0.006314375, 0.026033642, 0.022981694, -0.0452004, -0.0035584134, -0.01265506, -0.010938339, -0.011004114, -0.024481358, 0.0065511637, 0.05033083, 0.002900666, 0.019127294, 0.018232757, -0.008550717, -0.022100313, -0.028809335, 0.016575234, 0.004873908, -0.011076466, -0.0024731301, -0.007721954, 0.00066021393, -0.025033865, 0.0095110275, -0.0132799195, -0.0009759327, 0.0061466494, 0.006186114, 0.007123404, 0.011175129, 0.0015958596, 0.008169223, -0.006084163, -0.00596248, 0.06972122, 0.004189851, 0.015825402, 0.005248824, 0.017101433, -0.004804845, 0.012701102, 0.008070561, -0.03357143, 0.004239182, -0.026309896, -0.011431649, -0.00013802419, 0.013233878, 0.011543467, 0.029493393, -0.032071766, 0.015246584, -0.009011139, 0.004301668, -0.0102608595, -0.0055579655, 0.023573667, 0.0034169976, 0.015457063, -0.0065544527, -0.017798645, 0.00687346, -0.025770543, -0.0008830259, 0.00117819, 0.0004480904, 0.01382585, 0.016627854, -0.020705888, 0.008432322, 0.018548477, -0.003400554, 0.026257277, -0.022455497, 0.0023794011, 0.021705665, 0.0068076854, -0.0044101966, -0.007024742, 0.020995297, 0.02377099, -0.013733765, -0.009497873, -0.024441892, -0.00833366, 0.010024071, -0.011471114, 0.0112014385, -0.019311463, -0.004890352, -0.012911581, 0.010688395, 0.004337844, -0.024639217, 0.011530312, 0.008221842, -0.017982814, -0.012365651, -0.0012414982, -0.16417375, 0.006186114, 0.0166147, -0.034939542, 0.005564543, 0.024941782, 0.03343988, 0.0060151, 0.0089059, 0.0005113986, 0.03754422, 0.022218706, -0.023915695, -0.0020044853, -0.0061795367, 0.024284033, 0.0151676545, 0.045963388, -0.0050975424, 0.023021158, 0.002637567, -0.015825402, 0.011425072, -0.011497424, 0.008728308, 0.016259516, 0.013286497, -0.036333967, -0.031124607, 0.01135272, -0.0019469323, 0.0051304298, 0.023284258, 0.0006281488, 0.007176024, 0.01020824, 0.0073930807, 9.645249e-05, -0.016022727, 0.024560288, 0.0023826899, 0.030440548, -0.0050087464, 0.009767549, -0.016417375, 0.03017745, 0.005426416, -0.0010819945, 0.027204432, -0.0073601934, 0.02523119, -0.013760075, 0.0027526729, 0.037281122, 0.039596394, 0.019508788, -0.0023662462, 0.021126846, -0.0069852774, -0.024323499, -0.020377014, -0.017456615, 0.031966526, -0.01012931, 0.0056566275, -0.020034986, -0.017206673, 0.032334864, -0.02799373, 0.016785715, -0.029703872, 0.0049692816, 0.0030305712, -0.010958072, 0.004110921, 0.0113001, -0.023389498, 0.014352048, -0.00049906585, 0.0010244416, -0.013418047, 0.00440033, -0.031598184, -0.012747144, 0.002713208, 0.0071957563, 0.0117736785, 0.007051052, 0.004485837, -0.0144178225, 0.016706785, -0.007202334, 0.012477468, -0.016075347, 0.026559839, 0.011155396, -0.018680027, -0.018180138, -0.023257948, -0.006238734, 0.019127294, -0.0123261865, -0.009076914, 0.016601544, 0.011596086, 0.016825179, -0.022350257, 0.014207344, 0.040701408, -0.0136548355, 0.0042194496, -0.017561857, 0.013424624, 0.005663205, -0.0013228945, 0.026086261, -0.006001945, -0.010043803, 0.014812471, -0.012497201, 0.06745858, 0.025586374, 0.01986397, 0.0151545, -0.017219827, -0.02310009, -0.10839677, -0.01892997, 0.008438899, 0.03309785, -0.049409986, -0.004137231, -0.0085638715, 0.033860836, -0.014865091, 0.029730182, -0.028993504, -0.029203983, 0.0018532033, -0.008991407, 0.0048936405, -0.010135887, 0.01760132, -0.007248376, -0.017482925, 0.03830721, -0.011734214, -0.016035883, 0.016114812, -0.007136559, 0.020587493, 0.016851489, -0.03333464, 0.0062354454, -3.029235e-05, 0.0073141507, -0.0031703424, -0.0300459, 0.04848914, 0.01650946, -0.007721954, -0.022416031, 0.009655732, -0.02145572, 0.04788401, 0.0008805593, 0.029177673, 0.0002542605, -0.011865763, -0.036886476, -0.0078206165, -0.014391513, -0.039649013, 0.031492945, 0.022639666, -0.016009573, -0.033992387, -0.0013171391, -0.027730629, -0.005038345, 0.028467307, -0.009741239, 0.031545565, 0.0047226264, -0.01124748, 0.007149714, -0.009997761, 0.012457736, 0.013247033, -0.0071562915, 0.005334331, -0.029256603, -0.014352048, -0.056829374, 0.054303624, -0.01869318, 0.008767773, -0.006965545, -0.014575682, -0.01320099, -0.027072882, 0.027730629, -0.011411917, -0.025560064, 0.014457287, -0.038780786, -0.004716049, -0.01502295, -0.0041701184, -0.017338222, 0.018864196, 0.019403549, 0.008846703, -0.013878469, 0.023876231, -0.037412673, 0.012457736, 0.03025638, -0.001290007, 0.00040410354, -0.01393109, 0.012017045, -0.007873236, -0.0064064595, 0.0166147, 0.033308327, -0.030808887, -0.033360947, -0.06577474, 0.013339117, 0.0007350327, 0.0065676076, 0.027309671, -0.026888713, 0.021705665, -0.008524407, 0.0048245774, -0.010885719, 0.0020094183, -0.004341133, -0.0155359935, -0.0038576883, -0.009649155, -0.014049484, 0.03849138, -0.009333435, 0.030703647, 0.03280844, 0.0035945894, -0.012115707, 0.01382585, 0.039649013, -0.0029565745, -0.0069063473, -0.00010914496, 0.007892969, 0.0015687275, 0.023521047, -0.014904556, -0.037096955, -0.011457959, 0.005627029, -0.018153828, -0.039649013, -0.026783474, -0.003400554, 0.019824507, 0.028230518, -0.0031341664, -0.04764722, 0.024836542, -0.012490623, 0.018364308, 0.003288737, -0.024205104, -0.009708351, -0.0057026697, -0.009793859, 0.031545565, 0.008846703, 0.009392633, -0.025560064, -0.022376567, -0.012964201, 0.015878022, -6.8189904e-05, -0.022810679, -0.003824801, 0.02733598, 0.025967866, 0.010570001, 0.012102552, 0.0011995669, -0.008484942, -0.024928626, 0.02025862, 0.0050515, -0.015496528, -0.017193517, -0.005913149, 0.0079982085, 0.011359298, 0.0038379559, 0.0078206165, -0.028967194, -0.024481358, -0.024573442, 0.028256828, 0.029230293, 0.0017035657, -0.008471787, 0.028914575, 0.014838781, -0.0065347203, -0.032229625, 0.014299428, 0.016088502, 0.006429481, -0.014233653, 0.041464396, 0.007912701, -0.0076035596, 0.009017717, 0.023402652, 0.017509237, 0.016917264, 0.01312206, 0.036255036, 0.001963376, -0.003660364, -0.015601768, 0.001428134, -0.022245016, 0.009201886, -0.022008227, -0.04814711, -0.0056072967, -0.004390464, 0.012352496, -0.014667766, 0.0041569634, 0.0071299816, -0.003877421, 0.013214145, 0.024797076, -0.018811576, -0.011063311, 0.025691614, 0.012668215, 0.0027362292, -0.020245465, -0.00051386515, 0.031834975, -0.000110378234, 0.033913456, -0.04327978, 0.020995297, -0.032150693, -0.025849473, 0.00039074305, -0.034281794, 0.0001238415, 0.009563647, -0.008616491, -0.0067188893, 0.011069888, -0.001682189, 0.083086655, 0.0070181647, -0.008412589, -0.022692285, -0.011115931, 0.039070196, 0.007860081, 0.007886391, 0.00033257352, -0.02653353, -0.007425968, -0.03491323, -0.0069852774, -0.016049037, -0.019272, 0.041517016, -0.015956953, 0.023942005, 0.0043214005, 0.0005685404, 0.009215041, -0.016733095, -0.020021832, 0.0004320578, -0.036228728, 0.0004008148, 0.023376342, 0.005814487, 0.011319833, -0.037070643, -0.012253834, 0.0035814345, -0.046831615, -0.009800436, 0.0009290682, -0.018640561, 0.01929831, -0.024468202, 0.0030009726, 0.026691388, 0.010221395, -0.0022412743, -0.0020982143, -0.02759908, 0.01671994, -0.0028496906, -0.0021804327, -0.004265492, -0.006761643], \'node_index\': 0, \'node_id\': \'SMAD3_(144)\'}', "{'node_name': 'IL10RB', 'node_type': 'gene/protein', 'desc': 'IL10RB belongs to gene/protein node. IL10RB is interleukin 10 receptor subunit beta. The protein encoded by this gene belongs to the cytokine receptor family. It is an accessory chain essential for the active interleukin 10 receptor complex. Coexpression of this and IL10RA proteins has been shown to be required for IL10-induced signal transduction. This gene and three other interferon receptor genes, IFAR2, IFNAR1, and IFNGR2, form a class II cytokine receptor gene cluster located in a small region on chromosome 21. [provided by RefSeq, Jul 2008].', 'desc_emb': [-0.02927333, -0.006862564, -0.042112965, -0.020822087, -0.01688946, 0.017983304, -0.039117917, 0.016850395, -0.031461015, -0.012631285, 0.008887476, -0.024220815, 0.0005607574, 0.013620953, 0.0011622085, -0.008457752, 0.02084813, -0.0051436676, 0.0059184735, 0.019142257, -0.017175943, 0.017670777, 0.0033694278, -0.017800996, -0.012748483, 0.010352445, 0.028856626, -0.055108864, -0.0051632007, -0.026382457, 0.0072011347, 0.011895546, -0.02786696, 0.00757226, 0.010528241, -0.019350609, 0.00891352, -0.0072141564, 0.028023222, -0.0036884656, 0.038753305, 0.019142257, 0.003805663, 0.0034963917, 0.0126052415, -0.008294978, 0.0084772855, -0.021199724, -0.04562889, 0.0015398448, -0.009831567, 0.014311116, -0.023973398, -0.01761869, -0.021290878, -0.0016212319, 0.014141831, -0.0064751613, 0.006608636, -0.036617704, -0.01655089, -0.013816282, -0.017970283, -0.0014299721, -0.002070489, 0.01795726, 0.0033059458, 0.021290878, -0.017306162, 0.01735825, 0.034351885, 0.026486633, -0.014792928, 0.01360793, 0.017006658, -0.025483944, -0.0119281, 0.015782595, -0.00046878995, 0.00022442505, 0.021759667, -0.0018621379, 0.00041059815, 0.006458884, 0.024832847, -0.012136451, -0.003799152, 0.018946927, -0.019376652, -0.019793354, 0.033544526, 0.01119236, 0.023166038, 0.018660445, -0.017631711, 0.01005294, 0.011543953, 0.03148706, 0.0010433833, -0.024207793, -0.0072922884, 0.01152442, -0.010228736, 0.0033238512, -0.033831008, -0.0010181532, -0.019676156, -0.006830009, 0.026512677, -0.008021517, -0.03169541, 0.044587135, 0.020118903, -0.014011611, 0.023100927, -0.01614721, 0.015300783, -0.008731213, -0.013555843, -0.007337865, 0.01942874, 0.05159294, -0.0008260795, -0.028960802, 0.018972972, 0.004577213, -0.009701348, -0.024507297, -0.023283234, 0.0095711285, 0.010424065, 0.0028208785, 0.020809066, 0.0049353167, 0.006117058, 0.007298799, -0.0036168448, -0.01742336, 0.0014779905, -0.028908715, 0.0063026207, 0.027997179, -0.004892995, -0.009154426, 0.010658461, 0.0077936333, 0.018999016, 0.01629045, -0.009102338, 0.0022967453, -0.029195197, -0.01782704, 0.006006371, -0.02198104, -0.00033124568, 0.019337585, 0.005635246, 0.031174533, -0.0025929944, -0.0037372978, 0.003292924, -0.0016285568, -0.0005208777, 0.01461062, 0.026773116, 0.02017099, 2.9401106e-05, 0.0013835814, -0.010677993, 0.012826614, -0.004681389, 0.013217272, -0.009805524, 0.0060779923, -0.0022934899, 0.019220388, -0.0046065124, 0.007800144, -0.02606993, -0.008314511, -0.003906583, 0.0050232145, 0.029195197, -0.0024839358, -0.017983304, -0.010293846, 0.029637944, -0.010808213, 0.0012948695, 0.005889174, 0.028648276, 0.011641617, -0.0011076791, -0.03385705, -0.6279702, -0.011511398, 0.032372553, 0.012377357, -0.0045414027, 0.029611899, -0.00978599, -0.005352019, 0.0022316356, 0.019103192, 0.0004903576, 0.008503329, 0.008262423, -0.006745367, -0.044378784, -0.036279134, -0.008034539, -0.018907862, 0.012351314, -0.012553154, -0.02125181, 0.008607505, -0.014636665, 0.0044046724, -0.010645439, -0.0038447287, 0.02914311, -0.033205956, -2.7824231e-05, 0.01816561, -0.00038964098, 0.021994062, 0.014532489, 0.010065963, 0.056931935, 0.013295405, -0.011973677, 0.0059738164, 0.027111687, 0.02125181, -0.003095967, -0.021004396, 0.0004704177, -0.02198104, -0.028179485, 0.030888049, 0.009186981, 0.03799803, -0.0018263275, -0.023478564, 0.0037828747, 0.0035908008, -0.023205103, 0.0043590954, -0.0030048133, -0.003444304, 0.020184012, -0.001076752, 0.014519467, -0.018478138, -0.0072727553, 0.013790238, 0.00036705603, -0.015378915, 0.007162069, -0.022944665, -0.0010523358, -0.005869641, 0.01206483, 0.008685636, 0.02573136, 0.04997822, 0.0040172697, 0.018712532, 0.01293079, 0.021538295, 0.0007593421, 0.027189817, -0.0072857775, 0.01989753, 0.012364335, -0.021238789, -0.025015153, -0.0014202057, 0.020314232, -0.009213025, -0.0014275305, 0.013594909, 0.009831567, 0.00013316971, 0.0238562, -0.014871059, -0.03370079, -0.019845441, -0.014949191, -0.0070318496, 0.015053366, 0.028492013, 0.01734523, -0.009434398, -0.0039228606, -0.0005640129, -0.0020818831, -0.018686488, 0.012859169, -0.0031952593, -0.033726834, 0.0077610784, 0.015105454, -0.011804392, -0.008216847, -0.007936874, -0.013178207, 0.0063351756, 0.0032554858, -0.019077146, -0.01816561, 0.008112671, 0.013751172, -0.028492013, 0.020314232, -0.02458543, 0.0050362367, 0.003450815, -2.4670479e-05, 0.0055180485, -0.00048018416, -0.00653376, -0.0062309997, -0.014584577, 0.008809345, -0.0045739575, 0.008594482, -0.019324563, 0.019637091, 0.014467379, 0.048337452, 0.0016415787, -0.0026613597, -0.044743396, -0.018269787, -0.012748483, 0.0072076456, -0.02646059, 0.0003741774, -0.013256338, -0.018269787, 0.004798586, 0.015014301, 0.0038219404, -0.0023032562, 0.012735461, -0.0058305752, 0.0024774247, 0.013829304, -0.032138158, -0.01748847, -0.015118476, -0.0020037515, -0.006091014, -0.009141404, 0.0037372978, -0.021733623, 0.003139916, -0.010085495, -0.0006372614, -0.0043786285, 0.04187857, -0.009551596, -0.022059172, -1.7358352e-06, 0.0070969593, -0.041644175, -0.0040726126, 0.010808213, -0.0012395262, 0.011732771, -0.013503755, 0.00060999667, -0.0005868013, -0.019415718, 0.010782169, 0.010664972, -0.022788402, 0.036747925, 0.018686488, 0.017579624, 0.023491586, 0.0051762224, 0.028335748, -0.03932627, 0.024598451, -0.012273181, 0.0509158, 0.01052173, 0.0011361645, 0.02532768, 0.00021425166, -0.012514087, 0.0076829465, -0.0037763636, -0.012169006, 0.012566175, -0.027085641, -0.0021111825, -0.033232, -0.009213025, -0.03242464, 0.008125693, -0.014571555, -0.00405308, 0.014311116, 0.014180897, 0.022957686, 0.008047561, 0.009434398, -0.010430576, 0.0061007803, 0.010046429, -0.025379768, -0.0079303635, 0.015300783, 0.011876012, -0.023270212, 0.0002494516, -0.00445676, 0.0070904484, 0.0026939146, 0.014623643, -0.01976731, -0.007298799, -0.007826188, 0.02330928, 0.015652375, -0.011804392, -0.004111679, -0.0026385712, -0.024702627, 0.051384587, -0.008334043, 0.0005591297, -0.011185849, 0.01005294, 0.009753436, -0.005179478, -0.011075163, 0.0102678025, 0.017397316, -0.012741972, 0.02566625, -0.015352871, 0.040550333, -0.02084813, 0.01621232, 0.0027215863, -0.0009058389, 0.017983304, 0.011446288, 0.061932363, 0.013555843, 0.03377892, -0.0017384294, 0.0063091316, -0.008425198, 0.011387689, 0.018465117, -0.014011611, -0.010059452, 0.0052901646, 0.006110547, -0.01949385, 0.01782704, 0.008086626, -0.012006232, 0.015496112, -0.00097094866, 0.011426755, 0.014441336, 0.02358274, 0.0014023005, -0.008301489, -0.015300783, 0.025028175, 0.024937022, -0.0100203855, -0.0065174825, -0.039352313, -0.0009383938, -0.003779619, 0.016095122, 0.0021307154, 0.029038934, 0.024702627, 0.020561649, 0.017123856, 0.01642067, 0.03752924, -0.019585002, 0.00482463, 0.007962919, 0.011622084, -0.00904374, 0.0030243464, 0.007975941, 0.032815296, 0.011042608, 0.0069602286, -0.0079824515, -0.014558532, 0.008737724, 0.0022820956, -0.014415291, 0.005801276, -0.016850395, 0.016381605, 0.009186981, 0.013712106, 0.011889034, 0.0033922163, -0.015704464, -0.0025050964, -0.02935146, 0.012631285, 0.019611048, 0.00037031152, 0.028648276, -0.0030276019, 0.017097812, -0.013210761, -0.0031415438, -0.037893854, -0.008698658, 0.002773674, -0.017866107, -0.015170564, -0.0027525134, 0.0063156425, -0.005622224, -0.0007495756, 0.004551169, 0.030835962, -0.03966484, 0.007787122, 0.0013152163, 0.008679125, 0.022124281, 0.0024041764, 0.03357057, 0.016433693, -0.008633548, 0.00090828055, 0.012110407, 0.016329518, -0.045941416, 0.027893003, 0.016602978, -0.01199321, 0.0049353167, 0.0021176937, -0.012494555, 0.0031285218, 0.0047139437, -0.0027639074, 0.00633192, 0.020275166, 0.025900645, 0.013373536, -0.0015618193, -0.0067063007, 0.003095967, -0.02927333, 0.0036005673, 0.018347919, -0.009759947, 0.010293846, -0.023843179, -0.007279266, -0.025301635, 0.001671692, -0.009688326, -0.0092260465, -0.014219962, 0.0076699248, -0.0153398495, -0.012442467, -0.03268508, 0.033804964, 0.0023797601, -0.022749335, -0.030184865, 0.0030601567, -0.01842605, -0.024754714, -0.014180897, 0.010215715, -0.011485354, 0.015001278, -0.009675304, 0.019663135, -0.0013843954, 0.01983242, -0.0071230032, 0.021499228, 0.026981467, -0.01580864, -0.016824352, 0.027189817, -0.025015153, 7.380797e-05, 0.0134516675, -0.0020932774, 0.0038577507, 0.0036721881, -0.015730508, 0.003831707, -0.011218404, 0.013660018, 0.0053650406, 0.005794765, 0.027710695, 0.020392364, 0.02077, -0.011075163, -0.016095122, -0.026161084, 0.0009611822, -0.01734523, 0.007357398, -0.0030210908, -0.01185648, 0.0071230032, -0.00076259754, -0.012943812, -0.008640059, -0.0011915078, -0.007279266, 0.0032912963, -0.0029348205, 0.005166456, 0.017462427, 0.011237937, -0.0063579637, 0.014701774, -0.024194771, -0.020079836, -0.019728243, 0.011980188, 0.008607505, -0.008054072, 0.010879833, -0.0059803277, -0.0013705596, -0.0045739575, -0.004261431, -0.011361646, 0.03443002, 0.007585282, -0.025614163, -0.03349244, -0.0063221534, -0.007897808, 0.0046944106, -0.010274313, -0.010612884, -0.0009855983, 0.006973251, 0.009037228, 2.5331748e-05, -0.028283661, -0.020418407, -0.029455636, 0.028205529, 0.0022820956, 0.012142962, -0.012169006, 0.008666104, -0.016160231, -0.008301489, -0.018647423, -0.02922124, 0.0055766474, -0.014949191, 0.05797369, 0.03987319, 0.029794207, -0.009440909, 0.024910977, 0.0026385712, 0.022840489, 0.008308, 0.010209204, -0.002070489, -0.01989753, -0.0045381472, 0.0015740274, -0.008640059, 0.02351763, -0.025770426, 0.004730221, 0.007591793, -0.01634254, -0.0018784154, -0.0044404827, -0.024155704, -0.023166038, 0.0202361, -0.01052173, 0.00938231, -0.016694132, -0.0039619263, -0.003704743, -0.011283514, 0.0077741, 0.0011776721, 0.03008069, -0.022137303, 0.042243183, -0.0063612196, 0.008353577, -0.0067844326, 0.004534892, -0.00050093787, 0.012103897, 0.019064125, 0.014884082, 0.014024633, -0.004424205, -0.014818972, 0.0039814594, -0.009427887, -0.021681536, -0.010306868, -0.01574353, 0.01058684, 0.0045674467, -0.03432584, -0.00669979, -0.033362217, 0.0019093425, 0.016694132, -0.014818972, 0.015105454, -0.0012924279, -0.024338013, 0.044248562, 0.010437088, 0.020483516, -0.021928953, 0.028153442, -0.0014812461, 0.016173253, -0.014805949, -0.015587267, -0.0060942695, 0.0011263981, 0.023478564, 0.033232, 0.003102478, -0.01099052, 0.021186702, -0.004121445, -0.0072532226, -0.03411749, 0.017462427, -0.0076699248, 0.021915931, -0.0039619263, -0.0072076456, -0.018504182, -0.006745367, -0.0020753723, 0.008119182, -0.016446715, -0.0060193934, -0.015483091, 0.004909273, -0.029507725, 0.010691016, 0.00042280622, -0.016954571, -0.006185423, -0.013894414, 0.00055058405, -0.015014301, 0.01769682, 0.01774891, 0.0060812477, -0.008464263, 0.0003363324, -0.017735887, -0.017136877, -0.0067648995, -0.0024318479, 0.03565408, -0.012383868, -0.012370846, -3.687855e-05, 0.004899506, 0.013386558, -0.0019093425, -0.0030927116, -0.0072141564, -0.015053366, 0.0053585297, 0.021290878, 0.008933053, -0.017397316, 0.026721029, -0.022931643, -0.004977638, -0.04127956, -0.0077024796, 0.0024334756, -0.03987319, 0.008392642, -0.0032750187, 0.009753436, 0.0062668105, 0.027528388, 0.020926263, -0.002241402, 0.018999016, -0.050238658, -0.0005111113, -0.0094734635, 0.010430576, 0.013412601, 0.00938231, -0.005641757, -0.03596661, 0.010918899, 0.0062570437, -0.010886345, -0.0042842194, -0.0182177, -0.00669979, 0.009460442, -0.031200577, 0.020262145, -0.0020297954, 0.017944237, 0.007389953, -0.014324138, 0.025770426, -0.01849116, 0.00040693572, 0.032346506, 0.008939564, -0.0007939316, 0.003704743, 0.007331354, 0.0006913838, 0.010957966, -0.009506019, 0.004596746, 0.020275166, -0.015183586, 0.016629022, -0.011784859, -0.0088288775, -0.02217637, 0.01179137, -0.0006694093, -0.0056450125, -0.0038349624, 0.019936595, -0.0058989404, 0.01942874, -0.01769682, -0.005387829, 0.0005273887, -0.024233837, -0.026239216, 0.01052173, -0.016720176, 0.031122444, -0.008620527, 0.0017075023, -0.024923999, -0.006592359, -0.037216716, -0.011915078, 0.007077426, 0.0058826627, -0.01574353, 0.008822367, -0.012631285, 0.017058745, -0.017006658, 0.017071769, -0.015652375, -0.030445304, 0.017462427, -0.002107927, 0.027033554, 0.019676156, -0.02298373, -0.018204676, 0.039352313, 0.015157542, -0.018803686, 0.053702496, -0.044535045, 0.014936169, -0.018582312, 0.009929232, -0.029533768, 0.01393348, 0.011244448, -0.013178207, -0.014792928, 0.0098706335, 0.027684651, -0.0047627757, 0.0061268243, -0.009160937, 0.022410765, 0.004043313, 0.012051809, -0.029299373, -0.015040345, -0.00035566182, 0.023908287, 0.036227047, -0.0046878997, 0.006553293, 0.002260935, 0.0047269654, -0.0064263293, -0.018282808, 0.010150605, -0.007988962, 0.019259455, 0.007552727, -0.01125747, 0.0060421815, -0.021199724, -0.014766884, -0.005749188, -0.0005660476, -0.0020086349, -0.011277003, 0.0070904484, 0.027476301, -0.00978599, 0.0004553611, -0.020379342, -0.009069784, -0.028257618, 0.00884841, 0.0036005673, -0.015457047, -0.0254579, 0.0009383938, -0.028778495, -0.007389953, -0.00046187206, -0.018530225, -0.01467573, 0.013412601, 0.19553751, -0.021551317, -0.014415291, 0.005146923, -0.0021600148, 0.009538573, 0.0070253382, -0.0116611505, -0.02157736, -0.00034528496, 0.0036038228, 0.016563913, -0.00033083875, 0.0002494516, 0.007533194, 0.0019467805, -0.026434544, -5.6360597e-05, -0.047347788, 0.005436661, -0.002120949, 0.0028908714, 0.003597312, -0.008015006, 0.043519333, 0.010300357, -0.035029028, 0.010078984, 0.0059510283, 0.023843179, -0.028335748, 0.0052120327, -0.013425624, 0.009291156, 0.01574353, 0.0114528, -0.013972545, -0.015496112, 0.005286909, 0.010658461, -0.021290878, 0.01172626, -0.02453334, -0.022358676, 0.00435584, 0.021017417, 0.0035354577, 0.00019003899, -0.013959523, 0.0013412602, -0.009831567, -0.003538713, 0.026108997, 0.009903188, 0.00071173056, -0.004046569, 0.016720176, 0.009284645, 0.0023341833, 0.0009587406, 0.0039716926, 0.012175517, -0.0426078, 0.025848558, -0.012898235, 0.04252967, -0.011498376, -0.0011329091, 0.01842605, -0.02679916, 0.022632137, 0.0007255664, -0.0034898808, -0.007057893, -0.020066814, -0.015756551, 0.030445304, 0.02353065, 0.0134516675, 0.028309705, 0.013542821, 0.018139567, -0.04294637, -4.5831137e-05, 0.0047595203, -0.009128382, 0.008210335, 0.00053593435, -0.014897103, -0.010105028, 0.012299226, -0.012650818, -0.01588677, -0.008294978, 0.012142962, -0.014727818, -0.02653872, 0.0023472053, -0.023400431, -0.0182177, 0.0026629874, 0.07078728, 0.015795616, 0.0046944106, -0.027424213, 0.016954571, 0.0037893855, -0.0056938445, 0.004704177, 0.005407362, 0.028960802, -0.03805012, -0.003900072, -0.0054854937, 0.020288188, 0.0030113244, -0.0016765752, -0.030783875, 0.0068430314, -0.007787122, -0.014402269, -0.029169153, -0.009258602, 0.015248695, -0.013568865, 0.03515925, 0.0020916497, -0.009662282, 0.001336377, -0.037893854, 0.005146923, 0.008672615, 0.018243743, 0.00011750269, 0.025054218, -0.0025929944, 0.015170564, -0.011980188, -0.032007936, 0.027788827, -0.012598731, -0.010671482, 0.027241906, -0.015196608, -0.0030210908, 0.0043623513, 0.022072194, 0.029299373, -0.014910125, -0.009890166, -0.01842605, -0.05844248, -0.003939138, -0.00837311, 0.021017417, -0.020822087, -0.018959949, -0.021759667, 0.006058459, 0.0046325563, -0.02586158, 0.0050622807, 0.02995047, -0.027346082, -0.02211126, 0.0054854937, -0.1627222, 0.013660018, 0.025236527, -0.04435274, 0.025444878, 0.020548627, -0.009115361, -0.017254075, -0.0058240644, 0.00030255673, 0.034950897, 0.0027720463, -0.033388264, 0.0008281142, -0.01942874, 0.0051111127, -0.005039492, 0.04052429, 0.016811328, 0.022775378, 0.039274182, -0.023348344, 0.00326688, -0.022840489, 0.0092260465, 0.015756551, -0.008425198, -0.022228457, -0.0071490468, -0.0052706315, 0.008418687, 0.012032276, 0.04328494, -0.006566315, 0.0074876174, -0.016655065, 0.0325809, 0.0047204546, -0.0046195346, 0.024950044, 0.019871486, 0.024650538, -0.0030992224, 0.023283234, -0.020600714, 0.030132776, 0.025366746, 0.014897103, 0.025757404, -0.008854922, 0.006048693, -0.013555843, 0.018439071, 0.0074745957, 0.036227047, 0.015704464, 0.005505027, 0.013907435, -0.007891297, -0.00061894924, -0.0062961094, -0.020809066, 0.023543674, -0.029429592, 0.021824777, -0.0052543543, -0.0090762945, 0.029768163, -0.006149613, -0.002583228, -0.043988124, -0.015600288, 0.018321874, -0.008262423, 0.009616706, -0.001950036, -0.007077426, 0.026486633, 0.025054218, -0.014831994, -0.005462705, 0.019233411, -0.02646059, 0.00529342, 7.6605655e-05, 2.3602272e-05, 0.00864657, 0.023100927, -0.0070904484, -0.011049119, 0.021499228, -0.016824352, 0.011218404, -0.002824134, 0.014519467, 0.015248695, -0.0013274243, -0.0063026207, -0.025184438, 0.005319464, 0.029976513, -0.036227047, -0.015756551, -0.0076829465, 0.04042011, -0.013894414, -0.024246858, -0.004241898, 0.032867387, 0.0023651104, -0.002736236, -0.0048018415, -0.00163344, 0.010599862, -0.015652375, 0.013738151, -0.024598451, -0.009414865, -0.0023830156, -0.031148488, 0.06057808, 0.012338291, 0.008060583, 0.0033922163, -0.028283661, -0.035029028, -0.10761334, -0.043571424, 0.02097835, -0.0112118935, -0.024207793, 0.012715927, -0.020835109, 0.032711122, -0.01942874, 0.012950323, -0.027085641, -0.00015371997, -0.0026776372, -0.015300783, -0.017123856, 0.020197034, 0.0133019155, -0.0039814594, -0.0076764356, 0.033909142, 0.0017823785, -0.03268508, 0.004544658, -0.00034141907, 0.009056762, -0.010775658, -0.0347165, -0.015769573, 0.01748847, 0.014805949, 0.007136025, -0.024116639, 0.037294846, 0.005527815, -0.020210056, -0.0136990845, -0.019910552, -0.024494275, 0.03633122, -0.0048636957, 0.0030780619, -0.0027443746, -0.020965328, -0.021160658, 0.012071341, -0.0018458605, -0.02620015, -0.006745367, 0.035732213, -0.010918899, -0.02351763, -0.036669794, -0.029794207, -0.02807531, 0.01688946, -0.0009302551, 0.005902196, 0.027710695, -0.020509562, 0.0015984436, 0.013907435, -0.0044795484, -0.016160231, 0.007129514, 0.018178632, -0.03424771, 0.009063273, -0.024715649, 0.046931084, -0.017592646, -0.017188964, 0.010997031, 0.0013257966, -0.012461999, -0.0041084234, 0.0033059458, -0.02679916, -0.025926689, 0.025184438, -0.03385705, -0.02110857, -0.022879554, 0.009519041, -0.015105454, 0.02010588, 0.030914094, -0.008770279, -0.007643881, -0.0044925706, -0.018139567, -0.014858037, 0.03752924, 0.013946502, -0.009831567, 0.0020070071, 0.01983242, -0.021785712, -0.01742336, 0.020262145, 0.025223505, -0.02244983, -0.027971134, -0.07094355, 0.021316921, -0.025926689, -0.014884082, 0.0038577507, -0.015925838, 0.022475874, -0.0039554155, 1.4420785e-05, 0.015365893, -0.0076569025, 0.019754289, -0.011960655, 0.012598731, -0.00092781347, -0.013008921, 0.03344035, -0.017787974, 0.026877292, 0.014558532, -0.017280119, 0.005635246, 0.0127549935, 0.0066639795, -0.025288614, -0.004635812, -0.009532062, 0.024377078, -0.017384294, 0.0099227205, -0.019663135, -0.044066258, -0.017397316, 0.02453334, -0.012247138, -0.029038934, -0.0015618193, 0.028830582, 0.022098238, 0.026747072, -0.009877144, -0.055317216, 0.0191683, 0.010756126, 0.018126545, -0.00086351763, -0.024168726, 0.006419818, -0.0034247711, 0.0041377223, 0.0364354, 0.004450249, -0.0015097315, -0.03851891, -0.028882671, -0.0031057335, -0.011667661, -0.018178632, -0.016694132, -0.006686768, 0.019519893, 0.026512677, -0.0018133057, 0.030575523, -0.001741685, -0.0017433127, -0.0069927834, 0.010684504, 0.02351763, 0.0070709153, -0.022671204, 0.02558812, -0.0036070782, -0.0030373684, 0.022150327, 0.0030357405, -0.025002131, -0.012748483, -0.013047988, 0.010163627, 0.026213173, -0.009499508, -0.009252091, -0.0011280258, 0.0015846078, 0.011849969, -0.027788827, 0.017644733, 0.004590235, 0.009525551, -0.001741685, 0.01795726, -0.0008081744, 0.026174106, 0.0065630595, 0.015769573, -0.00037987452, 0.021785712, -0.0036168448, 0.05125437, 0.0020607226, -0.011543953, -0.010736592, -0.023322301, -0.0066477023, -0.011511398, -0.01186299, -0.039638795, -0.029299373, -0.019506872, 0.007188113, -0.008666104, 0.009317201, 0.0023390665, -0.015027323, -0.021147637, -0.018946927, -0.009186981, -0.020314232, 0.01580864, 0.015587267, 0.016694132, -0.0108668115, -0.0012167378, -0.001639951, -0.0053650406, 0.03544573, -0.013712106, 0.009265113, -0.0063742413, -0.023296257, -0.0074941283, -0.03229442, -0.008151736, -0.00047733562, 0.037190672, 0.021395054, 0.0050818133, -0.013633975, 0.08792416, 0.022150327, -0.009440909, 0.0020867665, -0.01836094, 0.021863842, 0.0053552743, 0.015600288, -0.023322301, -0.042842194, -0.0029022656, -0.027762784, 0.031929806, -0.025705317, -0.031461015, 0.010293846, 0.0012216211, 0.0063840076, -0.0076829465, -0.010502198, 0.0008142784, -0.021434119, -0.0079303635, 0.00055058405, -0.037216716, -0.00259625, 0.02238472, 0.0002734608, 0.0050232145, -0.050394922, 0.0095711285, -0.009720881, -0.050577227, -0.00036949763, 0.00817127, -0.03961275, -0.006680257, -0.010834257, 0.016915504, 0.03424771, 0.022163348, 0.011296536, -0.014871059, -0.018725555, 0.028570144, -0.008835388, -0.038310558, -0.0051925, -0.01776193], 'node_index': 1, 'node_id': 'IL10RB_(179)'}"]
CPU times: user 3.35 ms, sys: 154 μs, total: 3.5 ms
Wall time: 6.23 ms

In [50]:

%%time

# Assume node_coll_name is defined and collection exists
collection = Collection(node_coll_name)

# Load the collection into memory before query
collection.load()

# Vector similarity search in Milvus
vector_to_search = nodes_df["desc_emb"].iloc[0]
search_params = {"metric_type": "IP"}
results = collection.search(
    data=[vector_to_search],
    anns_field="desc_emb",
    param=search_params,
    limit=10,
    output_fields=["node_id", "node_name"]
)

results

%%time # Assume node_coll_name is defined and collection exists collection = Collection(node_coll_name) # Load the collection into memory before query collection.load() # Vector similarity search in Milvus vector_to_search = nodes_df["desc_emb"].iloc[0] search_params = {"metric_type": "IP"} results = collection.search( data=[vector_to_search], anns_field="desc_emb", param=search_params, limit=10, output_fields=["node_id", "node_name"] ) results

CPU times: user 5.9 ms, sys: 2.02 ms, total: 7.91 ms
Wall time: 11.6 ms

Out[50]:

data: [[{'node_index': 0, 'distance': 0.9999999403953552, 'entity': {'node_id': 'SMAD3_(144)', 'node_name': 'SMAD3'}}, {'node_index': 2182, 'distance': 0.8984372019767761, 'entity': {'node_id': 'SMAD protein signal transduction_(101792)', 'node_name': 'SMAD protein signal transduction'}}, {'node_index': 2023, 'distance': 0.8962712287902832, 'entity': {'node_id': 'SMAD protein complex_(55601)', 'node_name': 'SMAD protein complex'}}, {'node_index': 2913, 'distance': 0.8886438012123108, 'entity': {'node_id': 'heteromeric SMAD protein complex_(124869)', 'node_name': 'heteromeric SMAD protein complex'}}, {'node_index': 1009, 'distance': 0.8831610083580017, 'entity': {'node_id': 'regulation of SMAD protein signal transduction_(41433)', 'node_name': 'regulation of SMAD protein signal transduction'}}, {'node_index': 2152, 'distance': 0.8783092498779297, 'entity': {'node_id': 'positive regulation of SMAD protein signal transduction_(101088)', 'node_name': 'positive regulation of SMAD protein signal transduction'}}, {'node_index': 7, 'distance': 0.8766131401062012, 'entity': {'node_id': 'STAT3_(729)', 'node_name': 'STAT3'}}, {'node_index': 2223, 'distance': 0.8753368854522705, 'entity': {'node_id': 'SMAD protein complex assembly_(103465)', 'node_name': 'SMAD protein complex assembly'}}, {'node_index': 2858, 'distance': 0.8635619878768921, 'entity': {'node_id': 'co-SMAD binding_(123250)', 'node_name': 'co-SMAD binding'}}, {'node_index': 2149, 'distance': 0.860752284526825, 'entity': {'node_id': 'common-partner SMAD protein phosphorylation_(101002)', 'node_name': 'common-partner SMAD protein phosphorylation'}}]]

In [52]:

nodes_df.loc[0]

nodes_df.loc[0]

Out[52]:

	node_index	node_id	node_name	node_type	desc	desc_emb	feat	feat_emb
0	0	SMAD3_(144)	SMAD3	gene/protein	SMAD3 belongs to gene/protein node. SMAD3 is S...	[-0.03699171170592308, -0.005479035433381796, ...	MSSILPFTPPIVKRLLGWKKGEQNGQEEKWCEKAVKSLVKKLKKTG...	[-0.0010794274069904548, -0.0028632148270051, ...

Building Node Collection (Node Type-specific Embedding)¶

Note that nodes information of the PrimeKG data is different for each node type, we are going to build a separate collection for each node type.

We will use the node type as the collection name.

In [53]:

%%time

# Loop over group enrichment nodes by node_type
for node_type, nodes_df in tqdm(merged_nodes_df.groupby('node_type')):
    print(f"Processing node type: {node_type}")

    # Milvus collection name for this node_type
    node_coll_name = f"{milvus_database}_nodes_{node_type.replace('/', '_')}"

    # Define collection schema
    desc_emb_dim = len(nodes_df.iloc[0]['desc_emb'].to_arrow().to_pylist()[0])
    feat_emb_dim = len(nodes_df.iloc[0]['feat_emb'].to_arrow().to_pylist()[0])
    node_fields = [
        FieldSchema(name="node_index", dtype=DataType.INT64, is_primary=True, auto_id=False),
        FieldSchema(name="node_id", dtype=DataType.VARCHAR, max_length=1024),
        FieldSchema(name="node_name", dtype=DataType.VARCHAR, max_length=1024,
                    enable_analyzer=True, enable_match=True),
        FieldSchema(name="node_type", dtype=DataType.VARCHAR, max_length=1024,
                    enable_analyzer=True, enable_match=True),
        FieldSchema(name="desc", dtype=DataType.VARCHAR, max_length=40960,
                    enable_analyzer=True, enable_match=True),
        FieldSchema(name="desc_emb", dtype=DataType.FLOAT_VECTOR, dim=desc_emb_dim),        
        FieldSchema(name="feat", dtype=DataType.VARCHAR, max_length=40960,
                    enable_analyzer=True, enable_match=True),
        FieldSchema(name="feat_emb", dtype=DataType.FLOAT_VECTOR, dim=feat_emb_dim),
    ]
    schema = CollectionSchema(fields=node_fields, description=f"schema for collection {node_coll_name}")

    # Create collection if not exists
    if not utility.has_collection(node_coll_name):
        collection = Collection(name=node_coll_name, schema=schema)
    else:
        collection = Collection(name=node_coll_name)

    # Create index for node_index field (scalar)
    collection.create_index(
        field_name="node_index",
        index_params={"index_type": "STL_SORT"},
        index_name="node_index_index"
    )
    # Create index for node_name, node_type, desc fields (inverted)
    collection.create_index(
        field_name="node_name",
        index_params={"index_type": "INVERTED"},
        index_name="node_name_index"
    )
    collection.create_index(
        field_name="node_type",
        index_params={"index_type": "INVERTED"},
        index_name="node_type_index"
    )
    collection.create_index(
        field_name="desc",
        index_params={"index_type": "INVERTED"},
        index_name="desc_index"
    )
    collection.create_index(
        field_name="desc_emb",
        index_params={"index_type": "GPU_CAGRA", "metric_type": "IP"}, # AUTOINDEX
        index_name="desc_emb_index"
    )
    # Create index for feat_emb (vector)
    collection.create_index(
        field_name="feat_emb",
        index_params={"index_type": "GPU_CAGRA", "metric_type": "IP"}, # AUTOINDEX
        index_name="feat_emb_index"
    )

    # Prepare data for insertion
    # Normalize the embeddings
    graph_desc_emb_cp = cp.asarray(nodes_df["desc_emb"].list.leaves).astype(cp.float32).reshape(nodes_df.shape[0], -1)
    graph_desc_emb_norm = normalize_matrix(graph_desc_emb_cp, axis=1)
    graph_feat_emb_cp = cp.asarray(nodes_df["feat_emb"].list.leaves).astype(cp.float32).reshape(nodes_df.shape[0], -1)
    graph_feat_emb_norm = normalize_matrix(graph_feat_emb_cp, axis=1)
    # Columns must be lists of values in order matching schema fields
    data = [
        nodes_df["node_index"].to_arrow().to_pylist(),
        nodes_df["node_id"].to_arrow().to_pylist(),
        nodes_df["node_name"].to_arrow().to_pylist(),
        nodes_df["node_type"].to_arrow().to_pylist(),
        nodes_df["desc"].to_arrow().to_pylist(),
        graph_desc_emb_norm.tolist(), # Use normalized embeddings
        nodes_df["feat"].to_arrow().to_pylist(),
        graph_feat_emb_norm.tolist(), # Use normalized embeddings
    ]

    # Batch insert data in chunks
    batch_size = 500
    total_rows = len(data[0])
    for i in tqdm(range(0, total_rows, batch_size)):
        batch = [col[i:i + batch_size] for col in data]
        collection.insert(batch)

    # Flush the collection to ensure data is persisted
    collection.flush()

    # Print collection stats (number of entities and segment info)
    stats = collection.num_entities
    print(f"Collection {node_coll_name} stats:")
    print(stats)

%%time # Loop over group enrichment nodes by node_type for node_type, nodes_df in tqdm(merged_nodes_df.groupby('node_type')): print(f"Processing node type: {node_type}") # Milvus collection name for this node_type node_coll_name = f"{milvus_database}_nodes_{node_type.replace('/', '_')}" # Define collection schema desc_emb_dim = len(nodes_df.iloc[0]['desc_emb'].to_arrow().to_pylist()[0]) feat_emb_dim = len(nodes_df.iloc[0]['feat_emb'].to_arrow().to_pylist()[0]) node_fields = [ FieldSchema(name="node_index", dtype=DataType.INT64, is_primary=True, auto_id=False), FieldSchema(name="node_id", dtype=DataType.VARCHAR, max_length=1024), FieldSchema(name="node_name", dtype=DataType.VARCHAR, max_length=1024, enable_analyzer=True, enable_match=True), FieldSchema(name="node_type", dtype=DataType.VARCHAR, max_length=1024, enable_analyzer=True, enable_match=True), FieldSchema(name="desc", dtype=DataType.VARCHAR, max_length=40960, enable_analyzer=True, enable_match=True), FieldSchema(name="desc_emb", dtype=DataType.FLOAT_VECTOR, dim=desc_emb_dim), FieldSchema(name="feat", dtype=DataType.VARCHAR, max_length=40960, enable_analyzer=True, enable_match=True), FieldSchema(name="feat_emb", dtype=DataType.FLOAT_VECTOR, dim=feat_emb_dim), ] schema = CollectionSchema(fields=node_fields, description=f"schema for collection {node_coll_name}") # Create collection if not exists if not utility.has_collection(node_coll_name): collection = Collection(name=node_coll_name, schema=schema) else: collection = Collection(name=node_coll_name) # Create index for node_index field (scalar) collection.create_index( field_name="node_index", index_params={"index_type": "STL_SORT"}, index_name="node_index_index" ) # Create index for node_name, node_type, desc fields (inverted) collection.create_index( field_name="node_name", index_params={"index_type": "INVERTED"}, index_name="node_name_index" ) collection.create_index( field_name="node_type", index_params={"index_type": "INVERTED"}, index_name="node_type_index" ) collection.create_index( field_name="desc", index_params={"index_type": "INVERTED"}, index_name="desc_index" ) collection.create_index( field_name="desc_emb", index_params={"index_type": "GPU_CAGRA", "metric_type": "IP"}, # AUTOINDEX index_name="desc_emb_index" ) # Create index for feat_emb (vector) collection.create_index( field_name="feat_emb", index_params={"index_type": "GPU_CAGRA", "metric_type": "IP"}, # AUTOINDEX index_name="feat_emb_index" ) # Prepare data for insertion # Normalize the embeddings graph_desc_emb_cp = cp.asarray(nodes_df["desc_emb"].list.leaves).astype(cp.float32).reshape(nodes_df.shape[0], -1) graph_desc_emb_norm = normalize_matrix(graph_desc_emb_cp, axis=1) graph_feat_emb_cp = cp.asarray(nodes_df["feat_emb"].list.leaves).astype(cp.float32).reshape(nodes_df.shape[0], -1) graph_feat_emb_norm = normalize_matrix(graph_feat_emb_cp, axis=1) # Columns must be lists of values in order matching schema fields data = [ nodes_df["node_index"].to_arrow().to_pylist(), nodes_df["node_id"].to_arrow().to_pylist(), nodes_df["node_name"].to_arrow().to_pylist(), nodes_df["node_type"].to_arrow().to_pylist(), nodes_df["desc"].to_arrow().to_pylist(), graph_desc_emb_norm.tolist(), # Use normalized embeddings nodes_df["feat"].to_arrow().to_pylist(), graph_feat_emb_norm.tolist(), # Use normalized embeddings ] # Batch insert data in chunks batch_size = 500 total_rows = len(data[0]) for i in tqdm(range(0, total_rows, batch_size)): batch = [col[i:i + batch_size] for col in data] collection.insert(batch) # Flush the collection to ensure data is persisted collection.flush() # Print collection stats (number of entities and segment info) stats = collection.num_entities print(f"Collection {node_coll_name} stats:") print(stats)

  0%|          | 0/6 [00:00<?, ?it/s]

Processing node type: biological_process

100%|██████████| 4/4 [00:02<00:00,  1.67it/s]
 17%|█▋        | 1/6 [00:06<00:32,  6.52s/it]

Collection t2kg_primekg_nodes_biological_process stats:
1615
Processing node type: cellular_component

100%|██████████| 1/1 [00:00<00:00,  1.32it/s]
 33%|███▎      | 2/6 [00:11<00:22,  5.52s/it]

Collection t2kg_primekg_nodes_cellular_component stats:
202
Processing node type: disease

100%|██████████| 1/1 [00:00<00:00,  2.79it/s]
 50%|█████     | 3/6 [00:15<00:14,  5.00s/it]

Collection t2kg_primekg_nodes_disease stats:
7
Processing node type: drug

100%|██████████| 2/2 [00:01<00:00,  1.29it/s]
 67%|██████▋   | 4/6 [00:21<00:10,  5.49s/it]

Collection t2kg_primekg_nodes_drug stats:
748
Processing node type: gene/protein

100%|██████████| 1/1 [00:00<00:00,  2.78it/s]
 83%|████████▎ | 5/6 [00:26<00:05,  5.10s/it]

Collection t2kg_primekg_nodes_gene_protein stats:
102
Processing node type: molecular_function

100%|██████████| 1/1 [00:00<00:00,  1.27it/s]
100%|██████████| 6/6 [00:31<00:00,  5.20s/it]

Collection t2kg_primekg_nodes_molecular_function stats:
317
CPU times: user 1.77 s, sys: 238 ms, total: 2 s
Wall time: 31.2 s

In [54]:

# List all collections
for coll in utility.list_collections():
    print(f"Collection: {coll}")

    # Load the collection to get stats
    collection = Collection(name=coll)
    print(collection.num_entities)

# List all collections for coll in utility.list_collections(): print(f"Collection: {coll}") # Load the collection to get stats collection = Collection(name=coll) print(collection.num_entities)

Collection: t2kg_primekg_nodes
2991
Collection: t2kg_primekg_nodes_biological_process
1615
Collection: t2kg_primekg_nodes_cellular_component
202
Collection: t2kg_primekg_nodes_molecular_function
317
Collection: t2kg_primekg_nodes_disease
7
Collection: t2kg_primekg_nodes_gene_protein
102
Collection: t2kg_primekg_nodes_drug
748

In [62]:

merged_nodes_df[merged_nodes_df.node_type == 'gene/protein']

merged_nodes_df[merged_nodes_df.node_type == 'gene/protein']

Out[62]:

	node_index	node_id	node_name	node_type	desc	feat	desc_emb	feat_emb
48	16	IL1R2_(1654)	IL1R2	gene/protein	IL1R2 belongs to gene/protein node. IL1R2 is i...	MLRLYVLVMGVSAFTLQPAAHTGAARSCRFRGRHYKREFRLEGEPV...	[-0.02899833954870701, 0.0021472955122590065, ...	[-0.0014144123640997008, -0.001413213325978187...
49	17	HERC2_(1777)	HERC2	gene/protein	HERC2 belongs to gene/protein node. HERC2 is H...	MPSESFCLAAQARLDSKWLKTDIQLAFTRDGLCGLWNEMVKDGEIV...	[-0.014023775234818459, 0.009176542051136494, ...	[-0.007428297050351188, -0.010088199667370488,...
50	18	FCGR2A_(1990)	FCGR2A	gene/protein	FCGR2A belongs to gene/protein node. FCGR2A is...	MTMETQMSQNVCPRNLWLLQPLTVLLLLASADSQAAAPPKAVLKLE...	[-0.030694980174303055, 0.008019879460334778, ...	[0.005696470124890766, -0.006343481862099802, ...
51	19	CXCR1_(2012)	CXCR1	gene/protein	CXCR1 belongs to gene/protein node. CXCR1 is C...	MSNITDPQMWDFDDLNFTGMPPADEDYSPCMLETETLNKYVVIIAY...	[-0.027637897059321404, -0.0004926534020341933...	[0.010659271964210543, -0.0016738061039453167,...
52	20	FN1_(2057)	FN1	gene/protein	FN1 belongs to gene/protein node. FN1 is fibro...	MLRGPGPGLLLLAVQCLGTAVPSTGASKSKRQAQQMVQPQSPVAVS...	[-0.03269859775900841, 0.0019264572765678167, ...	[-0.0009853045367468072, -0.004182791243864574...
...	...	...	...	...	...	...	...	...
1168	848	IL17REL_(34781)	IL17REL	gene/protein	IL17REL belongs to gene/protein node. IL17REL ...	MSRSVLEALTSSTAMQCVPSDGCAMLLRVRASITLHERLRGLEACA...	[-0.02713669277727604, -0.011936459690332413, ...	[0.0003768516783435937, -0.005117051265290013,...
1169	849	TAGAP_(34814)	TAGAP	gene/protein	TAGAP belongs to gene/protein node. TAGAP is T...	MKLRSSHNASKTLNANNMETLIECQSEGDIKEHPLLASCESEDSIC...	[-0.03348768502473831, -0.009191920049488544, ...	[-0.0010800918479102228, -0.003681655277429923...
1170	850	DENND1B_(34887)	DENND1B	gene/protein	DENND1B belongs to gene/protein node. DENND1B ...	MDCRTKANPDRTFDLVLKVKCHASENEDPVVLWKFPEDFGDQEILQ...	[-0.02666034922003746, 0.005825655534863472, -...	[-0.000950883844682143, -0.0007957585650949377...
1171	851	IL21_(34967)	IL21	gene/protein	IL21 belongs to gene/protein node. IL21 is int...	MRSSPGNMERIVICLMVIFLGTLVHKSSSQGQDRHMIRMRQLIDIV...	[-0.02861925959587097, 0.005471138749271631, -...	[0.002036418016520262, -0.001478201039299714, ...
1172	852	FAM92B_(35156)	FAM92B	gene/protein	FAM92B belongs to gene/protein node.	MNIVFSRDSQVRVMENTVANTEKYFGQFCSLLAAYTRKTARLRDKA...	[-0.005422372370958328, -0.011813736520707607,...	[0.0008678497828269967, -0.0047449115179252635...

102 rows × 8 columns

In [76]:

%%time

# Assume node_coll_name is defined and collection exists
collection = Collection('t2kg_primekg_nodes_gene_protein')

# Load the collection into memory before query
collection.load()

# Vector similarity search in Milvus
vector_to_search = normalize_vector(merged_nodes_df[merged_nodes_df.node_type == 'gene/protein']['feat_emb'].iloc[0]).tolist()
search_params = {"metric_type": "IP"}
results = collection.search(
    data=[vector_to_search],
    anns_field="feat_emb",
    param=search_params,
    limit=10,
    output_fields=["node_id", "node_name"]
)

results

%%time # Assume node_coll_name is defined and collection exists collection = Collection('t2kg_primekg_nodes_gene_protein') # Load the collection into memory before query collection.load() # Vector similarity search in Milvus vector_to_search = normalize_vector(merged_nodes_df[merged_nodes_df.node_type == 'gene/protein']['feat_emb'].iloc[0]).tolist() search_params = {"metric_type": "IP"} results = collection.search( data=[vector_to_search], anns_field="feat_emb", param=search_params, limit=10, output_fields=["node_id", "node_name"] ) results

CPU times: user 8.64 ms, sys: 53 μs, total: 8.69 ms
Wall time: 20.8 ms

Out[76]:

data: [[{'node_index': 16, 'distance': 1.000000238418579, 'entity': {'node_id': 'IL1R2_(1654)', 'node_name': 'IL1R2'}}, {'node_index': 82, 'distance': 0.9777935147285461, 'entity': {'node_id': 'IL18RAP_(11588)', 'node_name': 'IL18RAP'}}, {'node_index': 59, 'distance': 0.9770528078079224, 'entity': {'node_id': 'IL12B_(6168)', 'node_name': 'IL12B'}}, {'node_index': 4, 'distance': 0.9742822051048279, 'entity': {'node_id': 'VCAM1_(417)', 'node_name': 'VCAM1'}}, {'node_index': 64, 'distance': 0.9739426970481873, 'entity': {'node_id': 'IL2RA_(7059)', 'node_name': 'IL2RA'}}, {'node_index': 51, 'distance': 0.9738389253616333, 'entity': {'node_id': 'ICAM1_(4968)', 'node_name': 'ICAM1'}}, {'node_index': 75, 'distance': 0.9732135534286499, 'entity': {'node_id': 'TLR9_(10113)', 'node_name': 'TLR9'}}, {'node_index': 18, 'distance': 0.9724373817443848, 'entity': {'node_id': 'FCGR2A_(1990)', 'node_name': 'FCGR2A'}}, {'node_index': 73, 'distance': 0.9714288711547852, 'entity': {'node_id': 'ICOSLG_(9454)', 'node_name': 'ICOSLG'}}, {'node_index': 31, 'distance': 0.9704809188842773, 'entity': {'node_id': 'TGFB1_(2889)', 'node_name': 'TGFB1'}}]]

In [80]:

# Check the ground truth for the search
merged_nodes_df[merged_nodes_df.node_type == 'gene/protein'].iloc[0]

# Check the ground truth for the search merged_nodes_df[merged_nodes_df.node_type == 'gene/protein'].iloc[0]

Out[80]:

	node_index	node_id	node_name	node_type	desc	feat	desc_emb	feat_emb
48	16	IL1R2_(1654)	IL1R2	gene/protein	IL1R2 belongs to gene/protein node. IL1R2 is i...	MLRLYVLVMGVSAFTLQPAAHTGAARSCRFRGRHYKREFRLEGEPV...	[-0.02899833954870701, 0.0021472955122590065, ...	[-0.0014144123640997008, -0.001413213325978187...

In [81]:

# Get node indices from the results
[n['node_index'] for n in results[0]]

# Get node indices from the results [n['node_index'] for n in results[0]]

Out[81]:

[16, 82, 59, 4, 64, 51, 75, 18, 73, 31]

In [82]:

# Get the cosine similarity scores
[n['distance'] for n in results[0]]

# Get the cosine similarity scores [n['distance'] for n in results[0]]

Out[82]:

[1.000000238418579,
 0.9777935147285461,
 0.9770528078079224,
 0.9742822051048279,
 0.9739426970481873,
 0.9738389253616333,
 0.9732135534286499,
 0.9724373817443848,
 0.9714288711547852,
 0.9704809188842773]

Building Edge Collection¶

Subsquently, we are also building the edges collection in Milvus.

Note that the edges information of PrimeKG has massive records, so once again we are chunking the data to avoid memory issues.

In [83]:

%%time

# Define collection name
edge_coll_name = f"{milvus_database}_edges"

# Define schema
edge_fields = [
    FieldSchema(name="triplet_index", dtype=DataType.INT64, is_primary=True, auto_id=False),
    FieldSchema(name="head_id", dtype=DataType.VARCHAR, max_length=1024),
    FieldSchema(name="head_index", dtype=DataType.INT64),
    FieldSchema(name="tail_id", dtype=DataType.VARCHAR, max_length=1024),
    FieldSchema(name="tail_index", dtype=DataType.INT64),
    FieldSchema(name="edge_type", dtype=DataType.VARCHAR, max_length=1024),
    FieldSchema(name="display_relation", dtype=DataType.VARCHAR, max_length=1024),
    FieldSchema(name="feat", dtype=DataType.VARCHAR, max_length=40960),
    FieldSchema(name="feat_emb", dtype=DataType.FLOAT_VECTOR, dim=1536),
]
edge_schema = CollectionSchema(fields=edge_fields, description="Schema for edges collection")

# Create collection if not exists
if not utility.has_collection(edge_coll_name):
    collection = Collection(name=edge_coll_name, schema=edge_schema)
else:
    collection = Collection(name=edge_coll_name)

# Create indexes
collection.create_index(field_name="triplet_index", index_params={"index_type": "STL_SORT"}, index_name="triplet_index_index")
collection.create_index(field_name="head_index", index_params={"index_type": "STL_SORT"}, index_name="head_index_index")
collection.create_index(field_name="tail_index", index_params={"index_type": "STL_SORT"}, index_name="tail_index_index")
collection.create_index(field_name="feat_emb", index_params={"index_type": "GPU_CAGRA", "metric_type": "IP"}, index_name="feat_emb_index") # AUTOINDEX

# Iterate over chunked edges embedding df
for edges_df in tqdm(edges_embedding_df):
    # Merge enrichment with embedding
    merged_edges_df = edges_enrichment_df.merge(
        edges_df[["triplet_index", "edge_emb"]],
        on="triplet_index",
        how="inner"
    )

    # Prepare data fields in column-wise format
    # Normalize the embeddings
    edges_edge_emb_cp = cp.asarray(merged_edges_df["edge_emb"].list.leaves).astype(cp.float32).reshape(merged_edges_df.shape[0], -1)
    edges_edge_emb_norm = normalize_matrix(edges_edge_emb_cp, axis=1)
    data = [
        merged_edges_df["triplet_index"].to_arrow().to_pylist(),
        merged_edges_df["head_id"].to_arrow().to_pylist(),
        merged_edges_df["head_index"].to_arrow().to_pylist(),
        merged_edges_df["tail_id"].to_arrow().to_pylist(),
        merged_edges_df["tail_index"].to_arrow().to_pylist(),
        merged_edges_df["edge_type_str"].to_arrow().to_pylist(),
        merged_edges_df["display_relation"].to_arrow().to_pylist(),
        merged_edges_df["feat"].to_arrow().to_pylist(),
        edges_edge_emb_norm.tolist(), # Use normalized embeddings
    ]

    # Insert in chunks
    batch_size = 500
    for i in tqdm(range(0, len(data[0]), batch_size)):
        batch_data = [d[i:i+batch_size] for d in data]
        collection.insert(batch_data)

    # Flush to ensure persistence
    collection.flush()

    # Print collection stats
    print(collection.num_entities)
    
    time.sleep(5)  # Sleep to avoid overwhelming the server

%%time # Define collection name edge_coll_name = f"{milvus_database}_edges" # Define schema edge_fields = [ FieldSchema(name="triplet_index", dtype=DataType.INT64, is_primary=True, auto_id=False), FieldSchema(name="head_id", dtype=DataType.VARCHAR, max_length=1024), FieldSchema(name="head_index", dtype=DataType.INT64), FieldSchema(name="tail_id", dtype=DataType.VARCHAR, max_length=1024), FieldSchema(name="tail_index", dtype=DataType.INT64), FieldSchema(name="edge_type", dtype=DataType.VARCHAR, max_length=1024), FieldSchema(name="display_relation", dtype=DataType.VARCHAR, max_length=1024), FieldSchema(name="feat", dtype=DataType.VARCHAR, max_length=40960), FieldSchema(name="feat_emb", dtype=DataType.FLOAT_VECTOR, dim=1536), ] edge_schema = CollectionSchema(fields=edge_fields, description="Schema for edges collection") # Create collection if not exists if not utility.has_collection(edge_coll_name): collection = Collection(name=edge_coll_name, schema=edge_schema) else: collection = Collection(name=edge_coll_name) # Create indexes collection.create_index(field_name="triplet_index", index_params={"index_type": "STL_SORT"}, index_name="triplet_index_index") collection.create_index(field_name="head_index", index_params={"index_type": "STL_SORT"}, index_name="head_index_index") collection.create_index(field_name="tail_index", index_params={"index_type": "STL_SORT"}, index_name="tail_index_index") collection.create_index(field_name="feat_emb", index_params={"index_type": "GPU_CAGRA", "metric_type": "IP"}, index_name="feat_emb_index") # AUTOINDEX # Iterate over chunked edges embedding df for edges_df in tqdm(edges_embedding_df): # Merge enrichment with embedding merged_edges_df = edges_enrichment_df.merge( edges_df[["triplet_index", "edge_emb"]], on="triplet_index", how="inner" ) # Prepare data fields in column-wise format # Normalize the embeddings edges_edge_emb_cp = cp.asarray(merged_edges_df["edge_emb"].list.leaves).astype(cp.float32).reshape(merged_edges_df.shape[0], -1) edges_edge_emb_norm = normalize_matrix(edges_edge_emb_cp, axis=1) data = [ merged_edges_df["triplet_index"].to_arrow().to_pylist(), merged_edges_df["head_id"].to_arrow().to_pylist(), merged_edges_df["head_index"].to_arrow().to_pylist(), merged_edges_df["tail_id"].to_arrow().to_pylist(), merged_edges_df["tail_index"].to_arrow().to_pylist(), merged_edges_df["edge_type_str"].to_arrow().to_pylist(), merged_edges_df["display_relation"].to_arrow().to_pylist(), merged_edges_df["feat"].to_arrow().to_pylist(), edges_edge_emb_norm.tolist(), # Use normalized embeddings ] # Insert in chunks batch_size = 500 for i in tqdm(range(0, len(data[0]), batch_size)): batch_data = [d[i:i+batch_size] for d in data] collection.insert(batch_data) # Flush to ensure persistence collection.flush() # Print collection stats print(collection.num_entities) time.sleep(5) # Sleep to avoid overwhelming the server

100%|██████████| 10/10 [00:06<00:00,  1.52it/s]

5000

100%|██████████| 9/9 [00:05<00:00,  1.62it/s]

9272

100%|██████████| 4/4 [00:02<00:00,  1.50it/s]

11272

100%|██████████| 3/3 [00:34<00:00, 11.59s/it]

CPU times: user 1.99 s, sys: 402 ms, total: 2.39 s
Wall time: 37 s

In [84]:

# List all collections
for coll in utility.list_collections():
    print(f"Collection: {coll}")

    # Load the collection to get stats
    collection = Collection(name=coll)
    print(collection.num_entities)

# List all collections for coll in utility.list_collections(): print(f"Collection: {coll}") # Load the collection to get stats collection = Collection(name=coll) print(collection.num_entities)

Collection: t2kg_primekg_edges
11272
Collection: t2kg_primekg_nodes_drug
748
Collection: t2kg_primekg_nodes
2991
Collection: t2kg_primekg_nodes_biological_process
1615
Collection: t2kg_primekg_nodes_cellular_component
202
Collection: t2kg_primekg_nodes_molecular_function
317
Collection: t2kg_primekg_nodes_disease
7
Collection: t2kg_primekg_nodes_gene_protein
102

In [85]:

%%time

# Assume node_coll_name is defined and collection exists
collection = Collection('t2kg_primekg_edges')

# Load the collection into memory before query
collection.load()

# Query by expr on triplet_index
expr = "triplet_index == 0"
output_fields = ["triplet_index", "head_id", "tail_id", "edge_type", "feat", "feat_emb"]

results = collection.query(expr, output_fields=output_fields)
results

%%time # Assume node_coll_name is defined and collection exists collection = Collection('t2kg_primekg_edges') # Load the collection into memory before query collection.load() # Query by expr on triplet_index expr = "triplet_index == 0" output_fields = ["triplet_index", "head_id", "tail_id", "edge_type", "feat", "feat_emb"] results = collection.query(expr, output_fields=output_fields) results

CPU times: user 10.6 ms, sys: 5.68 ms, total: 16.3 ms
Wall time: 1.52 s

Out[85]:

data: ["{'tail_id': 'LTF_(3233)', 'edge_type': 'drug|carrier|gene/protein', 'feat': 'Rose bengal (drug) has a direct relationship of drug_protein:carrier with LTF (gene/protein).', 'feat_emb': [-0.011582981, 0.0018133956, 0.0050554257, -0.022483429, -0.024731772, 0.0004805331, -0.0038308818, 0.02338009, -0.018321319, -0.0016402531, -0.0034862699, 0.010552491, -0.0005691955, 0.015858848, -0.0011275172, -0.0024306863, 0.028318414, -0.00040399912, -0.0013274257, -0.0055606337, -0.005219368, 0.027247775, 0.004904867, -0.012713844, -0.011161417, 0.01891017, 0.03642851, -0.03214595, -0.015577804, -0.012165141, 0.009816426, -0.0039312546, -0.02168045, -0.008250616, 0.008712329, -0.033591315, 0.010519033, 0.008431287, 0.028773436, -0.0077286786, 0.024785304, 0.010826842, 0.0058216024, 0.0050018937, -0.0031985354, -0.01072647, 0.021881195, -0.017317595, -0.040898427, 0.0023252952, -0.015136166, 0.011636513, -0.016086359, -0.005972161, -0.007989648, 0.005707847, 0.003285525, 0.0035431476, 0.0048078406, 0.009709362, -0.0065643582, -0.006213055, -0.016139891, 0.011429077, -0.0068721673, -0.0017364434, 0.0026247397, 0.02722101, 0.00970267, 0.015430591, 0.020609811, 0.0072134333, 0.014105675, 0.0036669401, 0.019927278, 0.01457408, -0.033189826, 0.016608296, -0.010525725, -0.007032763, 0.011415694, 0.0053900005, -0.025855944, 0.014547314, 0.018374851, 0.0015850483, -0.016420934, 0.02261726, -0.0011275172, -0.011415694, 0.01072647, 0.022804622, -0.0041085794, 0.003404299, -0.0031868254, -0.00044038412, 0.008103403, 0.037044127, -0.00044414809, -0.03308276, -0.006447257, 0.017893063, -0.025601666, -0.019954044, -0.028960798, -0.01082015, 0.004804495, -0.015497507, 0.015925763, -0.02467824, -0.018870022, 0.04349473, 0.013302696, -0.027836626, 0.012412727, 0.0037305094, 0.017277446, -0.015189698, 0.0038308818, -0.00875917, 0.030272331, 0.012051386, 0.008578499, 0.0053665806, 0.02257711, 0.0203823, -0.004710814, -0.0057111927, -0.024611326, 0.0053799637, 0.001222871, 0.0058416766, -0.008397829, 0.016139891, -0.0074141785, 0.017237296, -0.040737834, -0.023125812, -0.017826147, -0.019325044, 0.0087792445, 0.022563728, -0.029656714, 0.0097762775, -0.017973362, 0.022189004, 0.018642511, 0.012031311, -0.0150157185, -0.0055807084, 0.01274061, -0.015457357, -0.00011375544, -0.020971151, 0.00086236664, 0.002342024, 0.00010329998, 0.016996402, 0.0032955622, 0.013630579, -0.005410075, 0.01591238, 0.01457408, -0.00031721877, 0.039640427, 0.031771228, 0.014922038, -0.0073807207, 0.0036267913, -0.015069251, 0.021104982, -0.0051591443, -0.026658922, 0.027033648, 0.030325864, 0.0071331356, 0.010585948, -0.012017928, -0.007815668, -0.020087874, -0.0078022853, 0.0255749, 0.038409192, 0.011957705, -0.03067382, -0.02588271, 0.018200872, -0.02874667, -0.0055439053, -0.011723503, -0.0056074746, 0.024878984, 0.011723503, -0.014252888, -0.6616552, -0.037659746, 0.0048479894, -0.0020057762, 0.02722101, 0.023567451, -0.0028388675, 0.014226122, 0.0008042343, 0.019057384, -0.017852914, 0.019017234, 0.008518276, 0.009635756, 0.007494476, -0.016875956, -0.0022165584, -0.018896786, 0.033832207, 0.0077621364, -0.012760685, 0.0070662205, -0.017959978, 0.016661828, 0.02230945, 0.029656714, 0.020087874, -0.015136166, -0.004403005, 0.0040115523, -0.020168172, 0.03856979, 0.032386847, 0.012078152, 0.038944513, -0.012687078, -0.005794836, 0.023888644, 0.011248406, 0.008518276, -0.009682596, -0.008685563, 0.0016569819, -0.010646172, -0.017973362, 0.0164477, 0.016072975, 0.0029208383, -0.014346569, 0.0023805, -0.0036267913, 0.0053297775, 0.002765261, -0.022791239, -0.004884793, 0.014667761, 0.015109399, -0.030995013, 0.0020007577, -0.0059621236, -0.018374851, -0.009087053, -0.0026531785, -0.011208258, 0.0055807084, -0.0035264187, -0.012038003, 0.0006595307, 0.0153636765, 0.016661828, 0.021171896, 0.017491573, 0.0038108074, 0.017893063, -0.00849151, 0.035652295, 0.017812764, 0.011676662, -0.010646172, 0.01783953, -0.0032068999, 0.00239221, 0.011248406, 0.0076082316, 0.023273027, -0.0010823497, -0.025936242, 0.00097110344, 0.001997412, -0.023714665, 0.002872325, 0.026980115, -0.026725838, -0.023768196, -0.023848495, 0.0116030555, -0.013891547, -0.0016511268, 0.034073103, -0.030700587, 0.008116786, -0.014627612, 0.014293037, 0.00251433, 0.014975569, -0.0014077236, -0.021800896, 0.029121393, 0.013289313, -0.0034528123, -0.0090067545, -0.014694527, -0.00065158453, -0.006530901, -0.010753236, -0.020917619, -0.0030262293, 0.0067483746, 0.0028422133, -0.011469225, 0.010003787, 0.00042219163, -0.0009903415, -0.012452875, -0.00938817, 0.031583868, -0.028853733, -0.00587848, 0.0041989144, -0.0025310586, -0.008056562, -0.020850705, 0.017732468, -0.0025193486, 0.014667761, -0.01913768, 0.019699767, -0.00015045413, -0.002024178, -0.027622499, -0.018655894, 0.0010848589, 0.0097160535, -0.0024674896, -0.02714071, -0.016367402, -0.010920523, 0.008263999, -0.015229846, -0.010913831, -0.0027318036, -0.020248469, -0.012747302, 0.009695979, 0.03881068, -0.0035832964, -0.017183764, 0.009997097, -0.0047710375, -0.015229846, 0.021894578, -0.01063948, 0.004576984, -0.012867749, -0.0071732844, -0.010612714, 0.017197147, 0.039801024, -0.012586705, -0.013918313, -0.0028689792, -0.03843596, 0.002940913, 0.02914816, -0.008043179, 0.0026899818, -0.014640995, 0.00032934712, 0.020435832, -0.017090084, 0.007655072, 0.015537655, -0.0045702923, -0.0069993054, 0.03329689, 0.012861057, -0.013530207, 0.0068788584, 0.0033139638, 0.031717695, -0.050079163, 0.015591187, -0.01577855, -0.004004861, 0.0024290134, -0.008712329, 0.02655186, 0.0015666467, -0.0053130486, -0.0032754876, 0.00034858516, -0.025802411, 0.011301938, -0.029924374, -0.002062654, -0.010023862, -0.01747819, -0.009488543, 0.015711635, 0.006701534, 0.008785935, 0.015256613, -0.0090067545, -0.0021764094, -0.0002743514, 0.011368853, 0.014105675, -0.010345054, 0.020181555, -0.013985228, -0.0028087557, 0.004764346, 0.020636577, -0.0022182313, -0.010793384, -0.0039145257, -0.00391118, 0.007461019, 0.008464743, -0.010452118, -0.0073472634, 0.003324001, 0.026498327, 0.011034278, -0.015457357, -0.03583966, 0.014239505, -0.01694287, 0.030887948, -0.008919765, -0.00083350955, -0.0011843949, 0.024785304, 0.0027602424, 0.015497507, 0.018053658, 0.013971846, 0.03562553, -0.006848747, 0.0013851399, -0.009803043, 0.020542895, -0.026832903, -0.0137175685, -0.003867685, -0.004138691, 0.011288555, 0.032493908, 0.03522404, 6.2157735e-05, 0.007561391, 0.037044127, -0.0016737106, 0.0068520927, -0.019619469, 0.036589105, 0.007179976, -0.008819393, -0.016715359, -0.010017171, -0.026966732, 0.0011082792, 0.009287798, -0.033644848, 0.019391958, 0.0067115715, -0.0012755666, 0.01725068, 0.014493782, -0.002276782, -0.012446184, -0.017973362, 0.022416515, 0.042129666, -0.004379585, -0.0036100624, -0.013443217, 0.01810719, -0.007561391, 0.023099046, 0.04657282, 0.013122025, 0.034688722, 0.016795658, -0.010933906, -0.012539865, 0.0020041033, -0.018655894, 0.011322013, -0.017090084, -0.015617954, 0.010840225, -0.03723149, -0.0007013525, 0.043387666, -0.0131889405, -0.014480399, -0.0007682675, 0.013476674, -0.022068556, 0.03294893, 0.0032002083, -0.004991857, -0.01717038, 0.0076751467, -0.0009811408, -0.00061394484, 0.032788336, 0.010974055, -0.026966732, -0.00550041, -0.029549649, 0.0014821666, 0.013490058, 0.018562213, 0.032654505, -0.010659555, -0.010552491, 0.0007887602, -0.014199356, -0.003083107, -0.026431412, 0.0056442777, -0.00045083958, 0.0026899818, -0.024169687, 0.0021646994, -0.024624709, -0.007046146, 0.024517644, -0.006186289, -0.04868733, 0.022215769, 0.0038710309, 0.0007665946, 0.009020138, 0.015269996, 0.011168108, -0.0074141785, -0.020128023, 0.017906446, 0.028960798, 0.019338425, -0.011723503, 0.007775519, 0.019646235, -0.0049115587, -0.0025745535, 0.00077412254, -0.0033474213, -0.009615681, 0.007353955, 0.013316078, -0.0037238179, 0.0132491635, 0.018495297, 0.036134083, -0.0025260402, 0.010492267, -0.0039212173, -0.011214949, -0.006273278, -0.012580014, -0.0066814595, 0.009722745, -0.012820908, -0.01537706, -0.016661828, 0.01824102, -0.006079225, 0.010338362, -0.009220883, -0.010177767, -0.01430642, 0.0062431665, -0.014855123, 0.011429077, -0.0029810618, 0.015122782, 0.019873746, -0.026846286, -0.024102772, 0.0014821666, -0.011763652, -0.008317531, -0.009227574, -0.0021948111, 0.015082634, 0.029844075, 0.0112417145, 0.0054435325, -0.005624203, 0.010177767, 0.021961493, -0.027435137, -0.0078022853, 0.007494476, -0.023326557, -0.0037171263, -0.0016837479, 0.013784483, -0.017290829, -0.012493025, -0.00014229887, -0.0012680386, 0.0009384824, 0.020368917, -0.010378512, -0.01564472, 0.007989648, 0.008230542, 0.011522758, -0.019325044, -0.007099678, 0.020636577, -0.024397198, -0.023540685, 0.009950256, 0.010030554, -0.001768228, 0.016487848, -0.010612714, -0.024638092, 0.008926457, -0.004536835, -0.01877634, -0.0078022853, 0.0024139576, 0.011596364, 0.01162313, -0.020181555, 0.010753236, 0.02481207, -0.00938817, -0.00036008618, -0.01564472, -0.000392289, -0.0037137808, -0.019418724, 0.005450224, -0.0008544205, -0.03195859, -0.019673001, 0.015269996, 0.005968815, 0.011576289, 0.013851399, -0.01783953, -0.033109527, -0.007440944, -0.026324349, 0.013329461, -0.0050621172, -0.005657661, -0.028452244, 0.0010438735, 0.024116155, -0.017304212, 0.007815668, -0.025789028, -0.0056844265, 0.005637586, -0.0069256993, 0.0044933404, -0.010485576, 0.006497443, -0.036910295, 0.005550597, -0.018829873, -0.03482255, -0.0071866675, -0.009495234, 0.027863393, 0.021131746, 0.03755268, -0.023741432, 0.033055995, 0.0073472634, 0.021533238, 0.0342337, 0.0018886749, 0.01050565, -0.030914715, 0.014908655, 0.032199483, -0.009756203, 0.021626918, 0.004854681, -0.006253204, 0.031342972, -0.013784483, -0.007655072, -0.0035832964, -0.019793447, -0.033136293, -0.0010480557, -0.011288555, 0.0065476294, -0.04162111, -0.01622019, 0.019365191, -0.031985357, 0.00047551448, 0.012559939, 0.02284477, -0.040496938, 0.023366706, -6.743774e-06, -0.013690802, -0.021841045, -0.027970456, 0.02681952, -0.020823939, 0.01998081, -0.028559308, 0.014814974, -0.0042156433, -0.01940534, 0.006554321, 0.0020977843, -0.0035130358, 0.0009284452, -0.007340572, -0.012104917, -0.010164384, -0.03803447, -0.028050754, -0.030299097, -0.015002335, 0.02137264, -0.026297582, 0.027783094, -0.008933148, -0.01824102, 0.018522063, 0.009281106, 0.025066348, -0.03035263, 0.04437801, 0.034581657, 0.0021764094, -0.011094502, 0.004500032, 0.0011258443, -0.020101257, 0.0001032477, 0.022496812, -0.025146645, -0.0057747616, 0.013650653, 0.025414305, -0.028960798, -0.0056810807, -0.0020041033, -0.01591238, 0.0153636765, -0.01184395, 0.0030797615, -0.015818698, -0.002820466, -0.03329689, 0.0054468783, -0.008933148, 0.004275867, -0.0053531975, -1.09128905e-05, 0.01757187, 0.016862573, 0.0030764157, 0.0051056123, -0.01631387, 0.005169181, -0.009943564, -0.027287925, -0.01787968, 0.035705827, 0.01751834, 0.020261852, -0.004453191, -4.2788986e-05, -0.022336217, -0.0011935958, -0.0037271637, 0.03190506, -0.0018702734, -0.007835743, 0.011288555, 0.02874667, 9.498789e-05, 0.0049282876, -0.0052728998, -0.028291648, -0.019338425, -0.010438736, 0.018883405, 0.0131421, -0.020502746, 0.008832776, -0.020676725, -0.004540181, -0.010579256, -0.017290829, 0.01783953, -0.043414433, -0.010318289, 0.009287798, 0.0075346255, 0.019124297, 0.006898933, -0.006741683, 0.0007737043, 0.030807652, -0.04815201, 0.015069251, -0.009160659, 0.030272331, -0.01828117, 0.025762262, -0.020676725, -0.035866424, 0.020235086, -0.026806137, -0.030593524, 0.0028371946, -0.009602298, -0.005794836, 0.01877634, -0.03008497, 0.012774067, -0.00086738524, 0.021078216, -0.01457408, -0.0035331103, -0.010264756, -0.02181428, 0.006514172, 0.028398711, -0.027568966, -0.024745155, -0.011656588, 0.00827069, -0.0037003977, 0.0031416577, -0.0074810935, -0.0066814595, 0.03195859, -0.009153968, 0.017719084, 0.016099742, -0.01166997, -0.026136987, 0.026658922, 0.017090084, -0.01791983, 0.013389685, 0.011656588, -0.00083350955, 0.014868505, -0.022215769, -0.0047208513, 0.0058316393, -0.014426867, -0.013945079, 0.017812764, -0.0060357302, 0.038650088, -0.015443974, -0.051123034, -0.019525789, -0.010271448, -0.025387539, 0.0087792445, 0.0039346004, 0.008043179, -0.035652295, 0.0009225901, 0.0024976013, 0.019231362, 0.0021345876, 0.025896093, 0.013423143, 0.014587463, 0.008859542, -0.037900638, 0.02901433, 0.012038003, -0.021466322, -0.013684111, 0.014279654, -0.0010781675, -0.0006235639, 0.037713278, -0.029121393, 0.018428383, -0.0127205355, 0.016206807, -0.001970646, 0.010699703, 0.008170318, -0.010151001, -0.0029041097, 0.024611326, 0.027836626, -0.004958399, 0.004292595, 0.018120574, 0.026498327, 0.007634998, -0.008230542, -0.001781611, 0.011395619, 0.002340351, 0.023781579, 0.0009359731, -0.01725068, -0.0062164003, 0.017009785, -0.009060287, -0.009562149, -0.020542895, 0.0060491133, -0.015631337, -0.0019673002, 0.004707468, 0.008638723, 0.008424595, 0.005520485, -0.009588915, -0.0034795783, -0.017665552, -0.026257433, -0.004172148, 0.038409192, 0.0021513163, -0.014226122, -0.0024574522, -0.017344361, -0.007929424, -0.011134651, 0.002872325, 0.015965912, -0.017304212, -0.018040275, -0.0015858847, -0.0029743703, -0.019070767, -0.0131421, -0.0033557857, -0.017504957, 0.020877471, 0.19517758, -0.0109539805, 0.00978966, 0.016688593, -0.019726533, -0.009695979, -0.011181491, 0.001375939, 0.0057346127, 0.013202324, 0.0008816047, 0.026391264, -0.007407487, -0.0021111674, -0.0012672022, 0.0016854207, -0.03244038, -0.010887065, -0.015296762, 0.0007916877, -0.0009133893, -0.0010229626, -0.00040169893, -0.017893063, 0.033832207, -0.00489483, -0.015443974, -0.012319046, 0.0031617323, -0.011315322, -0.00018621181, 0.009401553, -0.014266271, -0.007829051, 0.0058048735, 0.008826084, -0.010719778, 0.0013458274, 0.016862573, -0.014640995, 0.0032621047, 0.017317595, 0.0055807084, -0.0236076, -0.01688934, 0.012278897, -0.0087190205, -0.007300423, -0.007815668, 0.00992349, -0.021854429, 0.016969636, 0.012031311, 0.009281106, -0.021841045, -0.011716811, 0.011910864, 0.008337606, -0.0008941513, 0.01877634, -0.008765861, 0.016916104, -0.017143615, 0.009301181, -0.020020958, 0.024129538, -0.026083454, 0.011944322, 0.04515422, -0.032654505, -0.009133893, -0.0068019065, 0.007902658, 0.021131746, -0.016916104, -0.016113125, 0.001668692, 0.012392652, 0.024290133, 0.028184583, 0.0010438735, -0.0073338803, -0.038248595, 0.0018836564, -0.021707216, -0.0022784548, -0.000107900385, 0.0069457735, -0.023072282, 0.0001611187, 0.021104982, -0.019325044, -0.008524967, -0.01680904, 0.018388234, -0.0038643393, 0.019432107, -0.003854302, -0.0025778993, -0.016340636, -0.052381035, 0.06306067, -0.0039011426, -0.0065743956, -0.017986745, 0.017732468, -0.012948046, 0.012934663, 0.007179976, -0.0063100816, 0.0036200997, -0.023928793, -0.002286819, -0.017090084, 0.0042892494, -0.0013541917, 0.0015833754, -0.025789028, 0.011409002, 0.0047442713, -0.00853835, -0.033216592, -0.00073313713, 0.02458456, -0.008424595, 0.01568487, -0.010880374, -0.004831261, -0.009314564, -0.045609243, 0.019391958, 0.011716811, -0.0052126762, 0.017223913, 0.008632031, 0.0103651285, 0.006932391, -0.021586768, -0.032788336, 0.008036488, -0.006008964, 0.0006913153, -0.00087324035, -0.02351392, -0.0031684237, -0.020971151, -0.010077395, 0.01680904, -0.018294552, -0.01654138, -0.016380785, -0.014493782, 0.0015725017, -0.016795658, 0.020971151, -0.030593524, -0.02261726, -0.03495638, 0.033216592, -0.007514551, -0.027261158, -0.008096712, -0.006795215, -0.02821135, 0.022523578, -0.016139891, -0.17162351, 0.0016427625, 0.0025344044, -0.022269301, -0.011536141, 0.0038911053, 0.002260053, -0.0044230795, -0.010599331, 0.026257433, 0.03915864, 0.002447415, -0.027354838, -0.013209014, -0.007494476, 0.005972161, -0.03875715, 0.028023988, 0.028853733, 0.0041319993, 0.013918313, -0.03396604, 0.00034879427, -0.023192728, 0.012934663, 0.0069725397, 0.0039847866, -0.0015172969, 0.006912316, -0.004229026, -0.050453886, 0.011168108, 0.0061059906, 0.00012598834, 0.02928199, 0.01417259, 0.0061762515, -0.008384446, -0.0030178651, 0.0026113566, 0.0072134333, 0.03683, -0.0102446815, 0.008056562, -0.020221703, 0.035786126, 0.023393473, -0.003456158, 0.022068556, -0.0009819772, 0.02400909, 0.0033474213, 0.01940534, 0.03589319, 0.031637397, 0.008183701, -0.0108603, 0.00862534, -0.010311597, 0.0026548514, -0.013222397, -0.028398711, 0.005132378, 0.0012989868, 0.0042624837, -0.019164447, -0.024290133, -0.026029922, -0.037285022, 0.004229026, -0.028157819, -0.01050565, -0.00021046848, -0.0015415535, 0.011248406, 0.0054903734, -0.013603813, 0.012325737, 0.01980683, -0.0019371883, -0.0045468723, 0.026297582, -0.014627612, 0.0045301435, -0.000635274, -0.0026314312, 0.027435137, 0.008979989, -0.015136166, -0.017598636, 0.03915864, -0.04009545, 0.017197147, -0.019472256, 0.026699072, 0.010840225, 0.0018418345, -0.0035331103, -0.0027384951, -0.006263241, 0.03294893, -0.010612714, -0.012171833, -0.0009092071, 0.03316306, 0.019472256, -0.025802411, 0.009341329, 0.031182375, -0.0097762775, -0.0061561773, 0.00039751673, 0.011522758, 0.038088, -0.017906446, 0.0131287165, 0.018950319, -0.020837322, 0.016233573, -0.023299793, 0.07034101, 0.01881649, 0.03236008, 0.04009545, 0.006992614, -0.0126000885, -0.09716053, -0.04140698, -0.005985544, 0.013449908, -0.013784483, 0.018174106, -0.0236879, 0.018843256, 0.021747366, 0.02821135, -0.00056459504, -0.034126636, -0.030486459, -0.0060491133, 0.022470046, 0.015818698, 0.02517341, 0.00037116898, -0.0033156367, 0.027729563, 0.0083108395, 2.1629741e-05, 0.015658103, 0.005654315, 0.015765168, -8.945956e-06, -0.041674644, -0.007929424, -0.006008964, 0.018950319, 0.01684919, -0.015885614, 0.0136573445, 0.00043118332, -0.01828117, -0.00728704, 0.012252131, 0.00478442, 0.025374155, -0.0057714162, -0.0025628433, -0.0076751467, 0.02521356, 0.008705637, -0.006597816, -0.01068632, -0.004777729, -0.009970331, 0.0111748, -0.027997222, -0.01654138, -0.03094148, -0.027729563, -0.0083576795, 0.019780064, -0.025775645, 0.009107128, 0.017692318, -0.019284895, -0.012486333, -0.02981731, 0.009803043, -0.014226122, 0.023928793, -0.010539108, -0.007855818, -0.0002816702, -0.035170507, 0.028586075, -0.015791934, -0.005065463, 0.014279654, -0.018388234, -0.010077395, -0.0537461, 0.004081813, -0.019726533, -0.00862534, 0.032386847, -0.053050186, -0.0036937061, -0.008043179, 0.030058203, -0.037713278, 0.016032826, 0.02177413, -0.012907897, 0.011616439, -0.01100082, -0.0020844014, 0.001927151, 0.014333186, -0.011830566, -0.01591238, 0.0042122975, 0.013971846, -0.023888644, 0.014828357, 0.014065526, 5.2996133e-06, -0.017264063, -0.017598636, -0.054468784, 0.025360772, 0.005824948, 0.011107884, 0.021934727, -0.006567704, 0.007969573, 0.0036903603, 0.008739095, 0.016246954, 0.0030396124, 0.011777034, 0.00013728024, -0.027435137, -0.0013173885, -0.028398711, 0.030727353, 0.021399407, 0.039265703, 0.03567906, 0.0020158135, 0.008645414, 0.0009719399, 0.012559939, -0.024758538, 0.0020844014, 0.0043896222, 0.02015479, -0.0120112365, -0.0017347705, -0.0019857017, -0.040389877, -0.015256613, -0.004734234, -0.029362287, -0.01444025, -0.0062966985, 0.022630643, 0.028371947, 0.008471435, -0.017411275, -0.010646172, 0.0111748, -0.016795658, -0.0004709141, -0.007775519, -0.03616085, -0.008886308, 0.003576605, 0.022362983, 0.009354712, 0.0062465123, -0.012211981, -0.016126508, -0.023982324, -0.012486333, 0.0064673317, -0.015269996, 0.0011400637, 0.0055439053, 0.010017171, 0.016126508, 0.011268481, 0.005727921, 0.0050621172, -0.004309324, -0.009040212, -0.0042992868, 0.014386718, -0.023085665, -0.040256046, 0.005637586, -0.0012337448, 0.0034896156, 0.005011931, 0.0030161922, -0.0017966669, -0.002554479, -0.026310965, 0.009642447, 0.014453633, -0.019499023, -0.011696736, 0.011736886, 0.005570671, 0.015551038, -0.0074810935, 0.0039981697, 0.022229152, 0.00077203143, 0.00066329463, 0.0132491635, 0.0005926157, 0.007815668, -0.004934979, 0.015725018, 0.010846917, -0.010746544, -0.0071063694, 0.03008497, -0.0035331103, -0.009829809, -0.02624405, -0.022684174, -0.0019238053, 0.023353323, -0.006320119, -0.034581657, -0.025829177, 0.0022918377, 0.0021345876, -0.010626097, -0.00849151, -0.0027803169, -0.019552555, 0.0044096964, -0.011094502, -0.005794836, 0.0026431412, 0.0137175685, 0.0013508459, 0.003563222, 0.011148034, 0.013985228, 0.019325044, 0.02592286, 0.029603181, -0.009107128, 0.027435137, -0.0124796415, -0.0062799696, 0.016755508, -0.023982324, 0.005242788, 0.008732404, 0.0022416515, 0.009762894, 0.007568083, -0.009421628, 0.08854189, 0.00036928698, -0.010512342, -0.008632031, 0.005517139, 0.01658153, -0.011469225, 0.030325864, -0.01654138, -0.019967427, 0.015403826, -0.014988952, 0.013757717, -0.035705827, -0.01810719, 0.008243924, 0.009622373, -0.013463291, -0.01810719, -0.008036488, 0.007942807, 0.0020024306, -0.0027368222, -0.0011643205, -0.017531723, 0.010980747, 0.034180168, 0.022470046, -0.0009744492, -0.021760749, -0.016354019, -0.0049650907, -0.02490575, 0.00094768323, 0.028960798, -0.012954738, -0.020810556, -0.020435832, 0.013075185, 0.01944549, -0.01332277, 0.0105457995, -0.024705006, -0.02391541, 0.005747996, -0.0046004043, -0.017558489, -0.021800896, -0.013336153], 'triplet_index': 0, 'head_id': 'Rose bengal_(14118)'}"]

In [86]:

# Check the ground truth for the search
results[0]['triplet_index'], results[0]['head_id'], results[0]['tail_id'], results[0]['edge_type'], results[0]['feat']

# Check the ground truth for the search results[0]['triplet_index'], results[0]['head_id'], results[0]['tail_id'], results[0]['edge_type'], results[0]['feat']

Out[86]:

(0,
 'Rose bengal_(14118)',
 'LTF_(3233)',
 'drug|carrier|gene/protein',
 'Rose bengal (drug) has a direct relationship of drug_protein:carrier with LTF (gene/protein).')

In [87]:

%%time

# Assume node_coll_name is defined and collection exists
collection = Collection('t2kg_primekg_edges')

# Load the collection into memory before query
collection.load()

# Vector similarity search in Milvus
vector_to_search = np.array(results[0]['feat_emb']).tolist() # merged_edges_df["edge_emb"].iloc[0]
search_params = {"metric_type": "IP"}
results = collection.search(
    data=[vector_to_search],
    anns_field="feat_emb",
    param=search_params,
    limit=10,
    output_fields=["head_id", "tail_id", "edge_type", "feat"]
)
results

%%time # Assume node_coll_name is defined and collection exists collection = Collection('t2kg_primekg_edges') # Load the collection into memory before query collection.load() # Vector similarity search in Milvus vector_to_search = np.array(results[0]['feat_emb']).tolist() # merged_edges_df["edge_emb"].iloc[0] search_params = {"metric_type": "IP"} results = collection.search( data=[vector_to_search], anns_field="feat_emb", param=search_params, limit=10, output_fields=["head_id", "tail_id", "edge_type", "feat"] ) results

CPU times: user 2.89 ms, sys: 0 ns, total: 2.89 ms
Wall time: 9.52 ms

Out[87]:

data: [[{'triplet_index': 0, 'distance': 1.000000238418579, 'entity': {'edge_type': 'drug|carrier|gene/protein', 'feat': 'Rose bengal (drug) has a direct relationship of drug_protein:carrier with LTF (gene/protein).', 'head_id': 'Rose bengal_(14118)', 'tail_id': 'LTF_(3233)'}}, {'triplet_index': 5636, 'distance': 0.9769061207771301, 'entity': {'edge_type': 'gene/protein|carrier|drug', 'feat': 'LTF (gene/protein) has a direct relationship of drug_protein:carrier with Rose bengal (drug).', 'head_id': 'LTF_(3233)', 'tail_id': 'Rose bengal_(14118)'}}, {'triplet_index': 196, 'distance': 0.900477409362793, 'entity': {'edge_type': 'drug|target|gene/protein', 'feat': '3h-Indole-5,6-Diol (drug) has a direct relationship of drug_protein:target with LTF (gene/protein).', 'head_id': '3h-Indole-5,6-Diol_(18278)', 'tail_id': 'LTF_(3233)'}}, {'triplet_index': 199, 'distance': 0.898104727268219, 'entity': {'edge_type': 'drug|target|gene/protein', 'feat': 'alpha-D-Fucopyranose (drug) has a direct relationship of drug_protein:target with LTF (gene/protein).', 'head_id': 'alpha-D-Fucopyranose_(18280)', 'tail_id': 'LTF_(3233)'}}, {'triplet_index': 198, 'distance': 0.8961387872695923, 'entity': {'edge_type': 'drug|target|gene/protein', 'feat': 'Nitrilotriacetic acid (drug) has a direct relationship of drug_protein:target with LTF (gene/protein).', 'head_id': 'Nitrilotriacetic acid_(18279)', 'tail_id': 'LTF_(3233)'}}, {'triplet_index': 197, 'distance': 0.8955156803131104, 'entity': {'edge_type': 'drug|target|gene/protein', 'feat': 'Lauric acid (drug) has a direct relationship of drug_protein:target with LTF (gene/protein).', 'head_id': 'Lauric acid_(14354)', 'tail_id': 'LTF_(3233)'}}, {'triplet_index': 201, 'distance': 0.8937671780586243, 'entity': {'edge_type': 'drug|target|gene/protein', 'feat': '(R)-Atenolol (drug) has a direct relationship of drug_protein:target with LTF (gene/protein).', 'head_id': '(R)-Atenolol_(18281)', 'tail_id': 'LTF_(3233)'}}, {'triplet_index': 5837, 'distance': 0.8916683197021484, 'entity': {'edge_type': 'gene/protein|target|drug', 'feat': 'LTF (gene/protein) has a direct relationship of drug_protein:target with (R)-Atenolol (drug).', 'head_id': 'LTF_(3233)', 'tail_id': '(R)-Atenolol_(18281)'}}, {'triplet_index': 202, 'distance': 0.8896037340164185, 'entity': {'edge_type': 'drug|target|gene/protein', 'feat': 'Parecoxib (drug) has a direct relationship of drug_protein:target with LTF (gene/protein).', 'head_id': 'Parecoxib_(15587)', 'tail_id': 'LTF_(3233)'}}, {'triplet_index': 5832, 'distance': 0.8886452913284302, 'entity': {'edge_type': 'gene/protein|target|drug', 'feat': 'LTF (gene/protein) has a direct relationship of drug_protein:target with 3h-Indole-5,6-Diol (drug).', 'head_id': 'LTF_(3233)', 'tail_id': '3h-Indole-5,6-Diol_(18278)'}}]]

In [88]:

# Get node indices from the results
[n['triplet_index'] for n in results[0]]

# Get node indices from the results [n['triplet_index'] for n in results[0]]

Out[88]:

[0, 5636, 196, 199, 198, 197, 201, 5837, 202, 5832]

In [89]:

# Get the cosine similarity scores
[n['distance'] for n in results[0]]

# Get the cosine similarity scores [n['distance'] for n in results[0]]

Out[89]:

[1.000000238418579,
 0.9769061207771301,
 0.900477409362793,
 0.898104727268219,
 0.8961387872695923,
 0.8955156803131104,
 0.8937671780586243,
 0.8916683197021484,
 0.8896037340164185,
 0.8886452913284302]