UMAP Visualization

In [ ]:

# import libraries
import pandas as pd
import numpy as np
import umap
import plotly.express as px
from sklearn.preprocessing import StandardScaler

# import libraries import pandas as pd import numpy as np import umap import plotly.express as px from sklearn.preprocessing import StandardScaler

In [ ]:

# Normalize the data
def is_data_normalized(data, tolerance=1e-3):
    """
    Check if the data is normalized (mean ~ 0 and std ~ 1 for each column).

    Parameters:
        data (np.ndarray): Data to check.
        tolerance (float): Tolerance level for mean and std deviation.

    Returns:
        bool: True if data is normalized, False otherwise.
    """
    mean_check = np.allclose(np.mean(data, axis=0), 0, atol=tolerance)
    std_check = np.allclose(np.std(data, axis=0), 1, atol=tolerance)
    return mean_check and std_check

# Normalize the data def is_data_normalized(data, tolerance=1e-3): """ Check if the data is normalized (mean ~ 0 and std ~ 1 for each column). Parameters: data (np.ndarray): Data to check. tolerance (float): Tolerance level for mean and std deviation. Returns: bool: True if data is normalized, False otherwise. """ mean_check = np.allclose(np.mean(data, axis=0), 0, atol=tolerance) std_check = np.allclose(np.std(data, axis=0), 1, atol=tolerance) return mean_check and std_check

In [ ]:

# Visualize single-cell data using UMAP
def visualize_umap(data, labels=None, n_neighbors=15, min_dist=0.1, n_components=2, random_state=42):
    """
    Visualize single-cell gene expression data on a UMAP plot.

    Parameters:
        data (pd.DataFrame or np.ndarray): Gene expression data. Rows are cells, columns are features/genes.
        labels (pd.Series or np.ndarray, optional): Labels or annotations for the cells (e.g., cell type, tissue).
        n_neighbors (int): Number of neighbors for UMAP.
        min_dist (float): Minimum distance between points on the UMAP plot.
        n_components (int): Number of dimensions for the UMAP embedding (default: 2 for 2D visualization).
        random_state (int): Random state for reproducibility.

    Returns:
        fig (plotly.graph_objects.Figure): Interactive UMAP plot.
    """
    # Ensure data is in the right format
    if isinstance(data, pd.DataFrame):
        data_array = data.values
    else:
        data_array = data

    # Check if data is normalized
    if not is_data_normalized(data_array):
        print("Data is not normalized. Performing normalization...")
        data_scaled = StandardScaler().fit_transform(data_array)
    else:
        print("Data is already normalized. Skipping normalization...")
        data_scaled = data_array

    # Apply UMAP
    umap_model = umap.UMAP(
        n_neighbors=n_neighbors,
        min_dist=min_dist,
        n_components=n_components,
        random_state=random_state
    )
    embedding = umap_model.fit_transform(data_scaled)

    # Create a DataFrame for the embedding
    umap_df = pd.DataFrame(embedding, columns=[f"UMAP_{i+1}" for i in range(n_components)])

    # Add labels if provided
    if labels is not None:
        umap_df["Labels"] = labels

    # Plot using Plotly
    if n_components == 2:
        fig = px.scatter(
            umap_df, x="UMAP_1", y="UMAP_2",
            color="Labels" if labels is not None else None,
            title="UMAP Visualization of Single-Cell Data",
            labels={"Labels": "Annotations"},
            hover_data=umap_df.columns
        )
    elif n_components == 3:
        fig = px.scatter_3d(
            umap_df, x="UMAP_1", y="UMAP_2", z="UMAP_3",
            color="Labels" if labels is not None else None,
            title="UMAP Visualization of Single-Cell Data (3D)",
            labels={"Labels": "Annotations"},
            hover_data=umap_df.columns
        )
    else:
        raise ValueError("Only 2D and 3D visualizations are supported.")

    # Customize plot appearance
    fig.update_traces(marker=dict(size=5, opacity=0.8), selector=dict(mode="markers"))
    fig.update_layout(legend_title="Cell Annotations", legend=dict(itemsizing="constant"))

    return fig

# Visualize single-cell data using UMAP def visualize_umap(data, labels=None, n_neighbors=15, min_dist=0.1, n_components=2, random_state=42): """ Visualize single-cell gene expression data on a UMAP plot. Parameters: data (pd.DataFrame or np.ndarray): Gene expression data. Rows are cells, columns are features/genes. labels (pd.Series or np.ndarray, optional): Labels or annotations for the cells (e.g., cell type, tissue). n_neighbors (int): Number of neighbors for UMAP. min_dist (float): Minimum distance between points on the UMAP plot. n_components (int): Number of dimensions for the UMAP embedding (default: 2 for 2D visualization). random_state (int): Random state for reproducibility. Returns: fig (plotly.graph_objects.Figure): Interactive UMAP plot. """ # Ensure data is in the right format if isinstance(data, pd.DataFrame): data_array = data.values else: data_array = data # Check if data is normalized if not is_data_normalized(data_array): print("Data is not normalized. Performing normalization...") data_scaled = StandardScaler().fit_transform(data_array) else: print("Data is already normalized. Skipping normalization...") data_scaled = data_array # Apply UMAP umap_model = umap.UMAP( n_neighbors=n_neighbors, min_dist=min_dist, n_components=n_components, random_state=random_state ) embedding = umap_model.fit_transform(data_scaled) # Create a DataFrame for the embedding umap_df = pd.DataFrame(embedding, columns=[f"UMAP_{i+1}" for i in range(n_components)]) # Add labels if provided if labels is not None: umap_df["Labels"] = labels # Plot using Plotly if n_components == 2: fig = px.scatter( umap_df, x="UMAP_1", y="UMAP_2", color="Labels" if labels is not None else None, title="UMAP Visualization of Single-Cell Data", labels={"Labels": "Annotations"}, hover_data=umap_df.columns ) elif n_components == 3: fig = px.scatter_3d( umap_df, x="UMAP_1", y="UMAP_2", z="UMAP_3", color="Labels" if labels is not None else None, title="UMAP Visualization of Single-Cell Data (3D)", labels={"Labels": "Annotations"}, hover_data=umap_df.columns ) else: raise ValueError("Only 2D and 3D visualizations are supported.") # Customize plot appearance fig.update_traces(marker=dict(size=5, opacity=0.8), selector=dict(mode="markers")) fig.update_layout(legend_title="Cell Annotations", legend=dict(itemsizing="constant")) return fig

In [ ]:

# Example usage
if __name__ == "__main__":
    # Simulated gene expression data (e.g., 100 cells with 50 genes each)
    data = np.random.rand(100, 50)
    labels = np.random.choice(["Type A", "Type B", "Type C"], size=100)

    # Create UMAP plot
    fig = visualize_umap(data, labels=labels, n_components=2)
    fig.show()

# Example usage if __name__ == "__main__": # Simulated gene expression data (e.g., 100 cells with 50 genes each) data = np.random.rand(100, 50) labels = np.random.choice(["Type A", "Type B", "Type C"], size=100) # Create UMAP plot fig = visualize_umap(data, labels=labels, n_components=2) fig.show()

/Users/jaydeepbhat/Documents/Hackathon/2024_AI_Agent/scripts/talk2cells/lib/python3.11/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

Data is not normalized. Performing normalization...

/Users/jaydeepbhat/Documents/Hackathon/2024_AI_Agent/scripts/talk2cells/lib/python3.11/site-packages/umap/umap_.py:1952: UserWarning: n_jobs value 1 overridden to 1 by setting random_state. Use no seed for parallelism.
  warn(