visualization-generator

name	visualization-generator
description	Data visualization skill for scientific figures using matplotlib/seaborn/plotly. Covers distribution, comparison, relationship, time series, and bioinformatics-specific charts (Circos, phylogenetic trees, heatmaps). Enforces publication-quality standards, English-only text, and proper file saving.

Visualization Generator

Expert-level data visualization skill for scientific figures and bioinformatics plots.

Critical Rules (MANDATORY)

import os
import matplotlib.pyplot as plt

# 1. ALWAYS save to results/
os.makedirs('results', exist_ok=True)
plt.savefig('results/plot.png', dpi=300, bbox_inches='tight')
plt.close()  # ALWAYS close after save

# 2. NEVER use plt.show() - headless environment

# 3. ALL text must be English
plt.title('Genome Size Distribution')  # ✅
# plt.title('')  # ❌ NEVER

# 4. Use the preferred color palette
COLORS = ['#ABD1BC', '#BED0F9', '#CCCC99', '#DBE4FB', 
          '#E3BBED', '#EDC3A5', '#F1F1F1', '#FCB6A5', '#FDEBAA']

Preferred Color Palette

Use these colors consistently across all figures:

Color	Hex	Use Case
Sage Green	#ABD1BC	Primary data
Soft Blue	#BED0F9	Secondary data
Olive	#CCCC99	Tertiary data
Light Periwinkle	#DBE4FB	Background/light
Lavender	#E3BBED	Categorical 5
Peach	#EDC3A5	Categorical 6
Light Gray	#F1F1F1	Neutral/grid
Coral	#FCB6A5	Highlight/warning
Cream	#FDEBAA	Accent

Supported Chart Types

Basic Charts

Category	Charts
Distribution	Histogram, KDE, Box, Violin, ECDF
Comparison	Bar, Grouped Bar, Stacked Bar
Relationship	Scatter, Heatmap, Pair Plot, Regression
Time Series	Line, Area, Multi-line
Statistical	Error Bars, QQ Plot, Residual Plot

Bioinformatics Charts

Type	Use Case
Circos/Chord	Genome comparisons, gene relationships
Phylogenetic Tree	Evolutionary relationships
Heatmap + Dendrogram	Gene expression, clustering
Volcano Plot	Differential expression
Manhattan Plot	GWAS results
Coverage Plot	Sequencing depth

Code Templates

Setup and Color Palette

import os
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns

# Setup
os.makedirs('results', exist_ok=True)
sns.set_style("whitegrid")
plt.rcParams.update({'savefig.dpi': 300, 'font.size': 10})

# Color palette
COLORS = ['#ABD1BC', '#BED0F9', '#CCCC99', '#DBE4FB', 
          '#E3BBED', '#EDC3A5', '#F1F1F1', '#FCB6A5', '#FDEBAA']

Distribution Plots

# Histogram with KDE
fig, ax = plt.subplots(figsize=(10, 6))
sns.histplot(data=df, x='value', kde=True, color=COLORS[0], ax=ax)
ax.set_xlabel('Value')
ax.set_ylabel('Frequency')
ax.set_title('Distribution of Values')
plt.tight_layout()
plt.savefig('results/distribution.png', dpi=300, bbox_inches='tight')
plt.close()

# Box + Violin comparison
fig, axes = plt.subplots(1, 2, figsize=(12, 5))
sns.boxplot(data=df, x='group', y='value', palette=COLORS[:3], ax=axes[0])
sns.violinplot(data=df, x='group', y='value', palette=COLORS[:3], ax=axes[1])
axes[0].set_title('Box Plot')
axes[1].set_title('Violin Plot')
plt.tight_layout()
plt.savefig('results/boxviolin.png', dpi=300, bbox_inches='tight')
plt.close()

Heatmap with Clustering

import scipy.cluster.hierarchy as sch

# Compute correlation and clustering
corr = df.select_dtypes(include=[np.number]).corr()

# Clustered heatmap
g = sns.clustermap(corr, cmap='RdBu_r', center=0,
                   figsize=(10, 10), annot=True, fmt='.2f',
                   dendrogram_ratio=0.15)
g.fig.suptitle('Clustered Correlation Heatmap', y=1.02)
plt.savefig('results/clustered_heatmap.png', dpi=300, bbox_inches='tight')
plt.close()

Volcano Plot (Differential Expression)

fig, ax = plt.subplots(figsize=(10, 8))

# Define significance thresholds
fc_thresh = 1.0  # log2 fold change
pval_thresh = 0.05

# Color by significance
colors = []
for _, row in df.iterrows():
    if row['padj'] < pval_thresh and abs(row['log2FC']) > fc_thresh:
        if row['log2FC'] > 0:
            colors.append('#FCB6A5')  # Coral - upregulated
        else:
            colors.append('#BED0F9')  # Blue - downregulated
    else:
        colors.append('#F1F1F1')  # Gray - not significant

ax.scatter(df['log2FC'], -np.log10(df['padj']), c=colors, alpha=0.7, s=20)
ax.axhline(-np.log10(pval_thresh), color='gray', linestyle='--', linewidth=1)
ax.axvline(-fc_thresh, color='gray', linestyle='--', linewidth=1)
ax.axvline(fc_thresh, color='gray', linestyle='--', linewidth=1)

ax.set_xlabel('Log2 Fold Change')
ax.set_ylabel('-Log10 Adjusted P-value')
ax.set_title('Volcano Plot')
plt.tight_layout()
plt.savefig('results/volcano.png', dpi=300, bbox_inches='tight')
plt.close()

Circos/Chord Diagram

import matplotlib.patches as mpatches
from matplotlib.path import Path
import matplotlib.patches as patches

# Data: connection matrix and labels
labels = ['Gene A', 'Gene B', 'Gene C', 'Gene D', 'Gene E']
matrix = np.array([
    [0, 15, 8, 3, 12],
    [15, 0, 10, 5, 2],
    [8, 10, 0, 20, 6],
    [3, 5, 20, 0, 9],
    [12, 2, 6, 9, 0]
])

n = len(labels)
colors_circos = COLORS[:n]

fig, ax = plt.subplots(figsize=(10, 10))
ax.set_aspect('equal')
ax.set_xlim(-1.5, 1.5)
ax.set_ylim(-1.5, 1.5)
ax.axis('off')

# Calculate segment angles
totals = matrix.sum(axis=1) + matrix.sum(axis=0)
gap = 0.03
total_arc = 2 * np.pi - n * gap
angles = totals / totals.sum() * total_arc

# Draw outer arcs and labels
start_angles = []
current = 0
for i in range(n):
    start_angles.append(current)
    theta1, theta2 = np.degrees(current), np.degrees(current + angles[i])
    wedge = mpatches.Wedge((0, 0), 1.0, theta1, theta2, width=0.12,
                           facecolor=colors_circos[i], edgecolor='white', linewidth=2)
    ax.add_patch(wedge)
    mid = current + angles[i] / 2
    ax.text(1.15 * np.cos(mid), 1.15 * np.sin(mid), labels[i],
            ha='center', va='center', fontsize=11, fontweight='bold')
    current += angles[i] + gap

# Draw chords
for i in range(n):
    for j in range(i + 1, n):
        if matrix[i, j] > 0:
            a1 = start_angles[i] + angles[i] / 2
            a2 = start_angles[j] + angles[j] / 2
            r = 0.88
            x1, y1 = r * np.cos(a1), r * np.sin(a1)
            x2, y2 = r * np.cos(a2), r * np.sin(a2)
            verts = [(x1, y1), (0, 0), (x2, y2)]
            codes = [Path.MOVETO, Path.CURVE3, Path.CURVE3]
            path = Path(verts, codes)
            lw = 1 + matrix[i, j] / matrix.max() * 5
            patch = patches.PathPatch(path, facecolor='none',
                                      edgecolor=colors_circos[i], alpha=0.6, linewidth=lw)
            ax.add_patch(patch)

ax.set_title('Chord Diagram', fontsize=14, fontweight='bold', pad=20)
plt.savefig('results/circos.png', dpi=300, bbox_inches='tight')
plt.close()

Phylogenetic Tree Visualization

from scipy.cluster.hierarchy import dendrogram, linkage
from scipy.spatial.distance import pdist

# Sample data - distance matrix or features
data = np.random.randn(10, 5)
labels = [f'Species_{i}' for i in range(10)]

# Compute linkage
Z = linkage(pdist(data), method='ward')

# Plot dendrogram
fig, ax = plt.subplots(figsize=(10, 8))
dendrogram(Z, labels=labels, orientation='left', ax=ax,
           leaf_font_size=10, color_threshold=0)
ax.set_xlabel('Distance')
ax.set_title('Phylogenetic Dendrogram')
plt.tight_layout()
plt.savefig('results/phylo_tree.png', dpi=300, bbox_inches='tight')
plt.close()

Sankey Diagram (Plotly)

import plotly.graph_objects as go

fig = go.Figure(go.Sankey(
    node=dict(
        pad=15, thickness=20,
        label=['Source A', 'Source B', 'Target X', 'Target Y', 'Target Z'],
        color=COLORS[:5]
    ),
    link=dict(
        source=[0, 0, 1, 1, 1],
        target=[2, 3, 2, 3, 4],
        value=[30, 20, 40, 25, 15]
    )
))
fig.update_layout(title='Flow Diagram', font_size=12)
fig.write_image('results/sankey.png', scale=2)

Figure Size Guidelines

Figure Type	Size (inches)
Single plot	(10, 6)
2x2 subplots	(12, 10)
Wide/timeline	(14, 6)
Square (radar, circos)	(10, 10)
Heatmap	(10, 8)

Quality Checklist

Before saving any figure:

• os.makedirs('results', exist_ok=True) called
• Save to results/descriptive_name.png
• plt.close() after save
• NO plt.show() anywhere
• ALL text in English
• Proper axis labels with units
• Descriptive title
• Legend if multiple series
• Color palette is accessible
• DPI ≥ 300 for publication

Output Format

When generating visualizations, provide:

{
  "code": "import os\nimport pandas as pd\n...",
  "description": "Brief description of the visualization",
  "has_visualization": true,
  "visualization_purpose": "WHY: Analysis goal, question being answered",
  "visualization_analysis": "WHAT: Key patterns, insights from the figure"
}