基于给定基因表达二维矩阵绘制3D图的技术咨询

Hey there! Let's walk through exactly how to visualize your cell-by-gene expression data in 3D. First, let's ground this: you have a matrix where each row is a cell, each column is a gene's expression value. Since 3D plots rely on three dimensions, we’ve got two straightforward approaches to choose from—let’s cover both with Python (the standard tool for bioinformatics visualization).

Option 1: Plot 3 Specific Genes Directly #

If you want to focus on the relationship between three particular genes, you can map each gene’s expression to an x/y/z axis. This is great if you have a hypothesis about how these genes correlate across cells.

Step-by-Step Code Example #

First, make sure your data is saved in a text/csv file (let’s call it gene_exp_data.txt). Here’s how to load it and plot:

import pandas as pd
import matplotlib.pyplot as plt
from mpl_toolkits.mplot3d import Axes3D

# Load your data (adjust the separator if your file uses tabs instead of spaces)
df = pd.read_csv("gene_exp_data.txt", sep="\s+", header=None)
# Assign column names to make it easier to reference
df.columns = ["Cell_ID", "Rpl37a", "Itm2c", "Atp1b1", "Olfm1", "Prnp"]

# Set up the 3D plot
fig = plt.figure(figsize=(10, 8))
ax = fig.add_subplot(111, projection="3d")

# Pick three genes for the axes—swap these out for any genes you care about!
x_vals = df["Rpl37a"]
y_vals = df["Itm2c"]
z_vals = df["Atp1b1"]

# Plot cells as scatter points. Optional: use a fourth gene's expression for color coding
scatter = ax.scatter(x_vals, y_vals, z_vals, c=df["Olfm1"], cmap="viridis", s=120)

# Add labels and title
ax.set_xlabel("Rpl37a Expression")
ax.set_ylabel("Itm2c Expression")
ax.set_zlabel("Atp1b1 Expression")
ax.set_title("3D Plot of Gene Expression (3 Target Genes)")

# Add a color bar to explain the fourth gene's values
plt.colorbar(scatter, label="Olfm1 Expression")

# Show the plot—you can drag to rotate it interactively!
plt.show()

Option 2: Reduce High-Dimensional Data to 3D (Recommended for Full Dataset) #

Since you have 5 genes (more than 3 dimensions), using a dimensionality reduction method like PCA (Principal Component Analysis) lets you capture the most meaningful variation in your data and plot it in 3D. This is ideal for spotting clusters of cells with similar expression profiles.

Step-by-Step Code Example #

We’ll use PCA here, but you can swap it for UMAP or t-SNE (popular for single-cell data) if you want better cluster separation.

import pandas as pd
import matplotlib.pyplot as plt
from mpl_toolkits.mplot3d import Axes3D
from sklearn.decomposition import PCA
from sklearn.preprocessing import StandardScaler

# Load and format data
df = pd.read_csv("gene_exp_data.txt", sep="\s+", header=None)
df.columns = ["Cell_ID", "Rpl37a", "Itm2c", "Atp1b1", "Olfm1", "Prnp"]

# Extract just the expression values (drop the cell ID column)
exp_data = df.drop("Cell_ID", axis=1)

# Standardize the data so all genes have equal weight in PCA
scaler = StandardScaler()
scaled_exp = scaler.fit_transform(exp_data)

# Run PCA to reduce to 3 dimensions
pca = PCA(n_components=3)
pca_results = pca.fit_transform(scaled_exp)

# Store results in a DataFrame for easy plotting
pca_df = pd.DataFrame(pca_results, columns=["PC1", "PC2", "PC3"])
pca_df["Cell_ID"] = df["Cell_ID"]

# Create the 3D plot
fig = plt.figure(figsize=(10, 8))
ax = fig.add_subplot(111, projection="3d")

# Plot cells, color-code by one gene's expression (e.g., Prnp)
scatter = ax.scatter(pca_df["PC1"], pca_df["PC2"], pca_df["PC3"], 
                     c=df["Prnp"], cmap="plasma", s=120)

# Optional: Label each cell with its ID (skip if you have hundreds of cells!)
for idx, cell in pca_df.iterrows():
    ax.text(cell["PC1"], cell["PC2"], cell["PC3"], cell["Cell_ID"], fontsize=8)

# Add labels showing how much variance each principal component explains
ax.set_xlabel(f"PC1 ({pca.explained_variance_ratio_[0]:.2%} variance)")
ax.set_ylabel(f"PC2 ({pca.explained_variance_ratio_[1]:.2%} variance)")
ax.set_zlabel(f"PC3 ({pca.explained_variance_ratio_[2]:.2%} variance)")
ax.set_title("3D PCA Plot of Cell Gene Expression")

# Add color bar
plt.colorbar(scatter, label="Prnp Expression")

plt.show()

Quick Tips #

• If you prefer R, you can use the rgl package for interactive 3D plots, or plot3D for static ones. The logic stays the same: either plot 3 genes directly or use PCA/UMAP to reduce dimensions.
• For large datasets, skip labeling individual cells—they’ll crowd the plot.
• Interactive 3D plots let you rotate the view, which is key for spotting patterns you might miss in a static image.

内容的提问来源于stack exchange，提问作者Nikita Vlasenko