Data Visualization (Universal)
Overview
This skill enables you to create professional scientific visualizations including scatter plots, line charts, heatmaps, violin plots, and more. Unlike cloud-hosted solutions, this skill uses the matplotlib and seaborn Python libraries and executes locally in your environment, making it compatible with ALL LLM providers including GPT, Gemini, Claude, DeepSeek, and Qwen.
When to Use This Skill
- Create publication-quality figures for papers and presentations
- Generate exploratory data analysis (EDA) plots
- Visualize gene expression, QC metrics, or clustering results
- Create multi-panel figures combining different plot types
- Export high-resolution images for reports
- Customize plot aesthetics (colors, fonts, styles)
How to Use
Step 1: Import Required Libraries
import matplotlib.pyplot as plt
import seaborn as sns
import pandas as pd
import numpy as np
from matplotlib import gridspec
import matplotlib.patches as mpatches
# Set style for publication-quality plots
sns.set_style("whitegrid")
plt.rcParams['figure.dpi'] = 150
plt.rcParams['savefig.dpi'] = 300
plt.rcParams['font.size'] = 10Step 2: Basic Scatter Plot
# Create figure and axis
fig, ax = plt.subplots(figsize=(6, 5))
# Scatter plot
ax.scatter(x_data, y_data, s=20, alpha=0.6, c='steelblue', edgecolors='k', linewidths=0.5)
# Labels and title
ax.set_xlabel('Gene Expression (log2)', fontsize=12)
ax.set_ylabel('Cell Count', fontsize=12)
ax.set_title('Expression vs. Cell Count', fontsize=14, fontweight='bold')
# Grid and styling
ax.grid(alpha=0.3)
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
# Save figure
plt.tight_layout()
plt.savefig('scatter_plot.png', dpi=300, bbox_inches='tight')
plt.show()
print("✅ Scatter plot saved to: scatter_plot.png")Step 3: Line Plot with Multiple Series
fig, ax = plt.subplots(figsize=(8, 5))
# Plot multiple lines
ax.plot(time_points, group1_values, marker='o', label='Group 1', color='#E74C3C', linewidth=2)
ax.plot(time_points, group2_values, marker='s', label='Group 2', color='#3498DB', linewidth=2)
ax.plot(time_points, group3_values, marker='^', label='Group 3', color='#2ECC71', linewidth=2)
# Styling
ax.set_xlabel('Time Point', fontsize=12)
ax.set_ylabel('Expression Level', fontsize=12)
ax.set_title('Gene Expression Over Time', fontsize=14, fontweight='bold')
ax.legend(frameon=True, loc='best', fontsize=10)
ax.grid(alpha=0.3, linestyle='--')
plt.tight_layout()
plt.savefig('line_plot.png', dpi=300, bbox_inches='tight')
plt.show()Step 4: Box Plot and Violin Plot
# Prepare data (long-form DataFrame)
# df should have columns: 'cluster', 'expression', 'gene', etc.
fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(12, 5))
# Box plot
sns.boxplot(data=df, x='cluster', y='expression', palette='Set2', ax=ax1)
ax1.set_title('Box Plot: Expression by Cluster', fontsize=12, fontweight='bold')
ax1.set_xlabel('Cluster', fontsize=11)
ax1.set_ylabel('Expression Level', fontsize=11)
ax1.tick_params(axis='x', rotation=45)
# Violin plot
sns.violinplot(data=df, x='cluster', y='expression', palette='muted', ax=ax2, inner='quartile')
ax2.set_title('Violin Plot: Expression Distribution', fontsize=12, fontweight='bold')
ax2.set_xlabel('Cluster', fontsize=11)
ax2.set_ylabel('Expression Level', fontsize=11)
ax2.tick_params(axis='x', rotation=45)
plt.tight_layout()
plt.savefig('box_violin_plot.png', dpi=300, bbox_inches='tight')
plt.show()Step 5: Heatmap
# Prepare data matrix (rows=genes, columns=samples or clusters)
# gene_expression_matrix: pandas DataFrame or numpy array
fig, ax = plt.subplots(figsize=(8, 6))
# Create heatmap
sns.heatmap(
gene_expression_matrix,
cmap='viridis',
cbar_kws={'label': 'Expression'},
xticklabels=True,
yticklabels=True,
linewidths=0.5,
linecolor='gray',
ax=ax
)
ax.set_title('Gene Expression Heatmap', fontsize=14, fontweight='bold')
ax.set_xlabel('Samples', fontsize=12)
ax.set_ylabel('Genes', fontsize=12)
plt.tight_layout()
plt.savefig('heatmap.png', dpi=300, bbox_inches='tight')
plt.show()Step 6: Bar Plot with Error Bars
fig, ax = plt.subplots(figsize=(7, 5))
# Data
categories = ['Cluster 0', 'Cluster 1', 'Cluster 2', 'Cluster 3']
means = [120, 85, 200, 150]
errors = [15, 10, 25, 20]
# Bar plot
bars = ax.bar(categories, means, yerr=errors, capsize=5,
color=['#E74C3C', '#3498DB', '#2ECC71', '#F39C12'],
edgecolor='black', linewidth=1.2, alpha=0.8)
# Labels
ax.set_ylabel('Cell Count', fontsize=12)
ax.set_title('Cell Counts by Cluster', fontsize=14, fontweight='bold')
ax.set_ylim(0, max(means) * 1.3)
# Add value labels on bars
for bar, mean in zip(bars, means):
height = bar.get_height()
ax.text(bar.get_x() + bar.get_width()/2., height + 5,
f'{mean}', ha='center', va='bottom', fontsize=10)
plt.tight_layout()
plt.savefig('bar_plot.png', dpi=300, bbox_inches='tight')
plt.show()Advanced Features
Multi-Panel Figure
# Create complex layout
fig = plt.figure(figsize=(12, 8))
gs = gridspec.GridSpec(2, 3, figure=fig, hspace=0.3, wspace=0.3)
# Panel A: Scatter
ax1 = fig.add_subplot(gs[0, :2])
ax1.scatter(x_data, y_data, c=cluster_labels, cmap='tab10', s=10, alpha=0.6)
ax1.set_title('A. UMAP Projection', fontsize=12, fontweight='bold', loc='left')
ax1.set_xlabel('UMAP1')
ax1.set_ylabel('UMAP2')
# Panel B: Violin
ax2 = fig.add_subplot(gs[0, 2])
sns.violinplot(data=df, y='expression', palette='Set2', ax=ax2)
ax2.set_title('B. Expression', fontsize=12, fontweight='bold', loc='left')
# Panel C: Heatmap
ax3 = fig.add_subplot(gs[1, :])
sns.heatmap(matrix, cmap='coolwarm', center=0, ax=ax3, cbar_kws={'label': 'Z-score'})
ax3.set_title('C. Gene Expression Heatmap', fontsize=12, fontweight='bold', loc='left')
plt.savefig('multi_panel_figure.png', dpi=300, bbox_inches='tight')
plt.show()Custom Color Palette
# Define custom colors
custom_palette = ['#E74C3C', '#3498DB', '#2ECC71', '#F39C12', '#9B59B6']
# Use in seaborn
sns.set_palette(custom_palette)
# Or create color dict for specific mapping
color_dict = {
'T cells': '#E74C3C',
'B cells': '#3498DB',
'Monocytes': '#2ECC71',
'NK cells': '#F39C12'
}
# Use in scatter plot
for cell_type, color in color_dict.items():
mask = df['celltype'] == cell_type
ax.scatter(df.loc[mask, 'x'], df.loc[mask, 'y'],
c=color, label=cell_type, s=20, alpha=0.7)
ax.legend()Density Plot
from scipy.stats import gaussian_kde
fig, ax = plt.subplots(figsize=(8, 6))
# Calculate density
xy = np.vstack([x_data, y_data])
z = gaussian_kde(xy)(xy)
# Sort points by density for better visualization
idx = z.argsort()
x, y, z = x_data[idx], y_data[idx], z[idx]
# Scatter with density colors
scatter = ax.scatter(x, y, c=z, s=20, cmap='viridis', alpha=0.6, edgecolors='none')
plt.colorbar(scatter, ax=ax, label='Density')
ax.set_xlabel('UMAP1', fontsize=12)
ax.set_ylabel('UMAP2', fontsize=12)
ax.set_title('Density Scatter Plot', fontsize=14, fontweight='bold')
plt.tight_layout()
plt.savefig('density_plot.png', dpi=300, bbox_inches='tight')
plt.show()Common Use Cases
QC Metrics Visualization
# Assuming adata.obs has QC columns: n_genes, n_counts, percent_mito
fig, axes = plt.subplots(1, 3, figsize=(15, 4))
# Plot 1: Histogram of genes per cell
axes[0].hist(adata.obs['n_genes'], bins=50, color='steelblue', edgecolor='black', alpha=0.7)
axes[0].axvline(adata.obs['n_genes'].median(), color='red', linestyle='--', label='Median')
axes[0].set_xlabel('Genes per Cell', fontsize=11)
axes[0].set_ylabel('Frequency', fontsize=11)
axes[0].set_title('Genes per Cell Distribution', fontsize=12, fontweight='bold')
axes[0].legend()
# Plot 2: Scatter UMI vs Genes
axes[1].scatter(adata.obs['n_counts'], adata.obs['n_genes'],
s=5, alpha=0.5, c='coral')
axes[1].set_xlabel('UMI Counts', fontsize=11)
axes[1].set_ylabel('Genes Detected', fontsize=11)
axes[1].set_title('UMIs vs Genes', fontsize=12, fontweight='bold')
# Plot 3: Violin plot of mitochondrial percentage
sns.violinplot(y=adata.obs['percent_mito'], ax=axes[2], color='lightgreen')
axes[2].axhline(y=20, color='red', linestyle='--', label='20% threshold')
axes[2].set_ylabel('Mitochondrial %', fontsize=11)
axes[2].set_title('Mitochondrial Content', fontsize=12, fontweight='bold')
axes[2].legend()
plt.tight_layout()
plt.savefig('qc_metrics.png', dpi=300, bbox_inches='tight')
plt.show()UMAP/tSNE Visualization
# Assuming adata.obsm['X_umap'] exists and adata.obs['clusters'] exists
fig, ax = plt.subplots(figsize=(8, 7))
# Get unique clusters
clusters = adata.obs['clusters'].unique()
n_clusters = len(clusters)
# Generate colors
colors = plt.cm.tab20(np.linspace(0, 1, n_clusters))
# Plot each cluster
for i, cluster in enumerate(clusters):
mask = adata.obs['clusters'] == cluster
ax.scatter(
adata.obsm['X_umap'][mask, 0],
adata.obsm['X_umap'][mask, 1],
c=[colors[i]],
label=f'Cluster {cluster}',
s=10,
alpha=0.7,
edgecolors='none'
)
ax.set_xlabel('UMAP1', fontsize=12)
ax.set_ylabel('UMAP2', fontsize=12)
ax.set_title('UMAP Projection by Cluster', fontsize=14, fontweight='bold')
ax.legend(bbox_to_anchor=(1.05, 1), loc='upper left', frameon=True, fontsize=9)
plt.tight_layout()
plt.savefig('umap_clusters.png', dpi=300, bbox_inches='tight')
plt.show()Gene Expression Dot Plot
# genes: list of gene names
# clusters: list of cluster IDs
# Create matrix: rows=genes, columns=clusters with mean expression and % expressing
fig, ax = plt.subplots(figsize=(10, 6))
# Prepare data
from matplotlib.colors import Normalize
# dot_size_matrix: % cells expressing (0-100)
# color_matrix: mean expression level
for i, gene in enumerate(genes):
for j, cluster in enumerate(clusters):
# Size proportional to % expressing
size = dot_size_matrix[i, j] * 5 # Scale factor
# Color by expression level
color_val = color_matrix[i, j]
ax.scatter(j, i, s=size, c=[color_val], cmap='Reds',
vmin=0, vmax=color_matrix.max(),
edgecolors='black', linewidths=0.5)
# Labels
ax.set_xticks(range(len(clusters)))
ax.set_xticklabels(clusters, rotation=45, ha='right')
ax.set_yticks(range(len(genes)))
ax.set_yticklabels(genes)
ax.set_xlabel('Cluster', fontsize=12)
ax.set_ylabel('Gene', fontsize=12)
ax.set_title('Marker Gene Expression', fontsize=14, fontweight='bold')
# Colorbar
norm = Normalize(vmin=0, vmax=color_matrix.max())
sm = plt.cm.ScalarMappable(cmap='Reds', norm=norm)
sm.set_array([])
cbar = plt.colorbar(sm, ax=ax, pad=0.02)
cbar.set_label('Mean Expression', rotation=270, labelpad=15)
plt.tight_layout()
plt.savefig('gene_dotplot.png', dpi=300, bbox_inches='tight')
plt.show()Volcano Plot (DEG Analysis)
# Assuming deg_df has columns: gene, log2FC, pvalue
fig, ax = plt.subplots(figsize=(8, 7))
# Calculate -log10(pvalue)
deg_df['-log10_pvalue'] = -np.log10(deg_df['pvalue'])
# Classify genes
deg_df['significant'] = 'Not Significant'
deg_df.loc[(deg_df['log2FC'] > 1) & (deg_df['pvalue'] < 0.05), 'significant'] = 'Up-regulated'
deg_df.loc[(deg_df['log2FC'] < -1) & (deg_df['pvalue'] < 0.05), 'significant'] = 'Down-regulated'
# Plot
for category, color in zip(['Not Significant', 'Up-regulated', 'Down-regulated'],
['gray', 'red', 'blue']):
mask = deg_df['significant'] == category
ax.scatter(deg_df.loc[mask, 'log2FC'],
deg_df.loc[mask, '-log10_pvalue'],
c=color, label=category, s=20, alpha=0.6, edgecolors='none')
# Threshold lines
ax.axvline(x=1, color='black', linestyle='--', linewidth=1, alpha=0.5)
ax.axvline(x=-1, color='black', linestyle='--', linewidth=1, alpha=0.5)
ax.axhline(y=-np.log10(0.05), color='black', linestyle='--', linewidth=1, alpha=0.5)
# Labels
ax.set_xlabel('log2 Fold Change', fontsize=12)
ax.set_ylabel('-log10(p-value)', fontsize=12)
ax.set_title('Volcano Plot: Differential Expression', fontsize=14, fontweight='bold')
ax.legend(frameon=True, loc='upper right')
plt.tight_layout()
plt.savefig('volcano_plot.png', dpi=300, bbox_inches='tight')
plt.show()Best Practices
- Figure Size: Use appropriate dimensions for target medium (papers: 6-8 inches wide, posters: larger)
- DPI: Save at 300 DPI for publications, 150 DPI for presentations
- Colors: Use colorblind-friendly palettes (e.g.,
viridis,Set2,tab10) - Fonts: Keep font sizes readable (titles: 12-14pt, labels: 10-12pt, ticks: 8-10pt)
- Transparency: Use alpha for overlapping points to show density
- Layout: Always call
plt.tight_layout()before saving to prevent label clipping - File Format: PNG for general use, SVG for vector graphics (editable in Illustrator)
- Close Figures: Call
plt.close()after saving to free memory when generating many plots
Defensive Validation Patterns
# Before plotting with AnnData color columns: verify they exist
if hasattr(data, 'obs'): # AnnData object
for col in color_columns:
assert col in data.obs.columns, f"Column '{col}' not found in adata.obs"
for basis in ['X_umap', 'X_pca']:
if basis in plot_args:
assert basis in data.obsm, f"Embedding '{basis}' not found. Compute it first."
# Before heatmap: verify data is numeric
if isinstance(data, pd.DataFrame):
non_numeric = data.select_dtypes(exclude='number').columns.tolist()
if non_numeric:
print(f"WARNING: Non-numeric columns will be skipped: {non_numeric}")Troubleshooting
Issue: "Figure too cluttered with many points"
Solution: Use transparency and smaller point sizes
ax.scatter(x, y, s=5, alpha=0.3, edgecolors='none')Issue: "Legend overlaps with data"
Solution: Place legend outside the plot area
ax.legend(bbox_to_anchor=(1.05, 1), loc='upper left')Issue: "Labels are cut off in saved figure"
Solution: Use bbox_inches='tight'
plt.savefig('plot.png', dpi=300, bbox_inches='tight')Issue: "Colors don't match between plots"
Solution: Define color palette once and reuse
PALETTE = {'Group A': '#E74C3C', 'Group B': '#3498DB'}
# Use PALETTE in all plotsIssue: "Heatmap text too small"
Solution: Adjust figure size or font size
fig, ax = plt.subplots(figsize=(12, 10))
sns.heatmap(data, ax=ax, annot_kws={'fontsize': 8})Technical Notes
- Libraries: Uses
matplotlibandseaborn(widely supported, stable) - Execution: Runs locally in the agent's sandbox
- Compatibility: Works with ALL LLM providers (GPT, Gemini, Claude, DeepSeek, Qwen, etc.)
- File Formats: Supports PNG, PDF, SVG, JPEG
- Performance: Typical plot generation takes <1 second for standard plots, 2-5 seconds for complex multi-panel figures
- Memory: Keep figure count reasonable; close figures after saving if generating many plots
References
- Matplotlib documentation: https://matplotlib.org/stable/contents.html
- Seaborn documentation: https://seaborn.pydata.org/
- Matplotlib gallery: https://matplotlib.org/stable/gallery/index.html
- Seaborn gallery: https://seaborn.pydata.org/examples/index.html