Scientific Charts

Scientific Charts

Generate publication-quality charts for articles. All output is dark-themed, high-DPI PNG ready for Substack embedding. Never generate charts via LLM — always use this skill's Python scripts.

Prerequisites

pip install --break-system-packages matplotlib numpy  # already done on this system

Chart Types

1. Comparison Bar Chart (Horizontal)

Use for: comparing treatments, compounds, protocols.

import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
import numpy as np

labels = ['Placebo', 'SSRI', 'Ketamine IV', 'Psilocybin']
values = [30, 47, 71, 68]
colors = ['#64748b', '#64748b', '#6366f1', '#10b981']

fig, ax = plt.subplots(figsize=(10, 4))
fig.patch.set_facecolor('#1a1a2e')
ax.set_facecolor('#1a1a2e')

bars = ax.barh(labels, values, color=colors, height=0.6)
ax.set_xlabel('Response Rate (%)', color='#94a3b8', fontsize=11)
ax.set_title('Treatment Response Comparison', color='#e2e8f0', fontsize=14, fontweight='bold')
ax.tick_params(colors='#94a3b8', labelsize=10)
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
ax.spines['left'].set_color('#334155')
ax.spines['bottom'].set_color('#334155')

for bar, val in zip(bars, values):
    ax.text(val + 1, bar.get_y() + bar.get_height()/2, f'{val}%',
            va='center', color='#e2e8f0', fontsize=10, fontweight='bold')

plt.tight_layout()
plt.savefig('chart.png', dpi=150, bbox_inches='tight', facecolor='#1a1a2e')

2. Dose-Response Curve

Use for: showing effect vs dose, receptor occupancy, concentration-response.

import numpy as np
import matplotlib.pyplot as plt

doses = np.logspace(-1, 2, 50)
response = 100 / (1 + np.exp(-(np.log10(doses) - 0.8) * 3))

fig, ax = plt.subplots(figsize=(8, 5))
fig.patch.set_facecolor('#1a1a2e')
ax.set_facecolor('#1a1a2e')

ax.plot(doses, response, color='#6366f1', linewidth=2.5)
ax.fill_between(doses, response - 8, response + 8, color='#6366f1', alpha=0.1)
ax.set_xscale('log')
ax.set_xlabel('Dose (mg/kg)', color='#94a3b8', fontsize=11)
ax.set_ylabel('Response (%)', color='#94a3b8', fontsize=11)
ax.set_title('Dose-Response Relationship', color='#e2e8f0', fontsize=14, fontweight='bold')
ax.tick_params(colors='#94a3b8')
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
ax.spines['left'].set_color('#334155')
ax.spines['bottom'].set_color('#334155')
ax.grid(True, alpha=0.1, color='#94a3b8')

# Annotate key points
ax.annotate('ED50 ≈ 6.3 mg/kg', xy=(6.3, 50), xytext=(15, 35),
            arrowprops=dict(arrowstyle='->', color='#f59e0b'),
            color='#f59e0b', fontsize=9)

plt.tight_layout()
plt.savefig('dose_response.png', dpi=150, bbox_inches='tight', facecolor='#1a1a2e')

3. Timeline / Mechanism Sequence

Use for: showing order of molecular events after administration.

import matplotlib.pyplot as plt

events = [
    ('NMDAR blockade', 0, 5),
    ('Glutamate surge', 5, 30),
    ('AMPA activation', 10, 60),
    ('BDNF release', 30, 120),
    ('mTORC1 activation', 45, 180),
    ('Spine formation', 120, 480),
    ('Clinical effect', 110, 1440),
]
labels, starts, ends = zip(*events)

fig, ax = plt.subplots(figsize=(10, 4))
fig.patch.set_facecolor('#1a1a2e')
ax.set_facecolor('#1a1a2e')

colors = ['#6366f1', '#818cf8', '#a78bfa', '#c4b5fd', '#ec4899', '#f472b6', '#10b981']
for i, (label, start, end) in enumerate(events):
    ax.barh(i, end - start, left=start, height=0.5, color=colors[i], edgecolor=None)
    ax.text(end + 5, i, f'{start}-{end}min', va='center', color='#94a3b8', fontsize=8)

ax.set_yticks(range(len(events)))
ax.set_yticklabels(labels, color='#e2e8f0', fontsize=10)
ax.set_xlabel('Time after administration (minutes)', color='#94a3b8', fontsize=11)
ax.set_xscale('log')
ax.set_title('Molecular Cascade After Ketamine Administration', color='#e2e8f0', 
             fontsize=13, fontweight='bold')
ax.tick_params(colors='#94a3b8')
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
ax.spines['left'].set_color('#334155')
ax.spines['bottom'].set_color('#334155')

plt.tight_layout()
plt.savefig('timeline.png', dpi=150, bbox_inches='tight', facecolor='#1a1a2e')

4. Forest Plot (Effect Size)

Use for: meta-analysis summaries, comparing multiple studies.

import matplotlib.pyplot as plt
import numpy as np

studies = ['Xu et al. (2016)', 'Caddy et al. (2014)', 'Fond et al. (2014)',
           'McGirr et al. (2015)', 'Kishimoto et al. (2016)']
effects = [0.91, 0.82, 0.78, 0.95, 0.71]
cis = [(0.70, 1.12), (0.55, 1.09), (0.52, 1.04), (0.68, 1.22), (0.48, 0.94)]

fig, ax = plt.subplots(figsize=(9, 5))
fig.patch.set_facecolor('#1a1a2e')
ax.set_facecolor('#1a1a2e')

y_positions = range(len(studies))
for i, (effect, (lo, hi)) in enumerate(zip(effects, cis)):
    color = '#6366f1' if lo > 0.5 else '#64748b'
    ax.errorbar(effect, i, xerr=[[effect - lo], [hi - effect]], 
                fmt='o', color=color, capsize=3, capthick=1.5, markersize=8)
    ax.text(hi + 0.03, i, f'd={effect:.2f}', va='center', color='#e2e8f0', fontsize=9)

ax.axvline(x=0, color='#475569', linewidth=1)
ax.axvline(x=0.5, color='#f59e0b', linestyle='--', alpha=0.5)
ax.axvline(x=0.8, color='#10b981', linestyle='--', alpha=0.5)
ax.set_yticks(y_positions)
ax.set_yticklabels(studies, color='#e2e8f0', fontsize=10)
ax.set_xlabel("Cohen's d (effect size)", color='#94a3b8', fontsize=11)
ax.set_title('Ketamine Meta-Analysis: Effect Sizes', color='#e2e8f0', 
             fontsize=14, fontweight='bold')
ax.tick_params(colors='#94a3b8')
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
ax.spines['left'].set_color('#334155')
ax.spines['bottom'].set_color('#334155')

plt.tight_layout()
plt.savefig('forest_plot.png', dpi=150, bbox_inches='tight', facecolor='#1a1a2e')

Design Standards

Never Do

• Pie charts with more than 5 slices
• 3D charts (they distort data)
• Dual y-axes without clear labeling on both sides
• Rainbow/disco color schemes
• Charts without source attribution in caption

PITFALLS

1. matplotlib needs Agg backend in headless mode. Always set matplotlib.use('Agg') before importing pyplot.
2. DPI matters for Substack. Substack compresses images. Generate at 150 DPI minimum, upload at 2x the intended display width.
3. Font rendering differs on WSL. If fonts look wrong, test with fc-list | grep -i dejavu to verify DejaVu fonts are installed.