Data Visualization Fundamentals

1. Start with the question

Use a distribution to inspect spread, a scatter plot to compare measurements, PCA to summarize major variation, a heatmap to inspect patterns, and UMAP to visualize local neighborhoods. No plot proves a mechanism by itself.

import seaborn as sns
import matplotlib.pyplot as plt

# Always label units and groups
sns.boxplot(data=df, x="batch", y="pct_mt", hue="condition", showfliers=False)
sns.stripplot(data=df, x="batch", y="pct_mt", hue="condition", dodge=True, color="black", alpha=0.55)
plt.xlabel("Batch")
plt.ylabel("Mitochondrial reads (%)")
plt.legend(title="Condition")

library(ggplot2)

# Always label units and groups
ggplot(df, aes(x = batch, y = pct_mt, color = condition)) +
  geom_boxplot(outlier.shape = NA) + geom_jitter(width = .15) +
  labs(x="Batch", y="Mitochondrial reads (%)")

2. PCA and UMAP

PCA axes are linear summaries; UMAP is a neighborhood visualization whose geometry depends on parameters. Report preprocessing, dimensions, neighbors, metric, and random seed.

sc.tl.pca(adata, n_comps=30, random_state=7)
sc.pp.neighbors(adata, n_neighbors=15, n_pcs=20, random_state=7)
sc.tl.umap(adata, random_state=7)

3. Heatmaps and volcano plots

Cluster and scale deliberately. A volcano plot combines effect size and significance; a small p-value alone is not a large or meaningful biological effect.

Figure checklist: title, units, n, group definition, preprocessing, color meaning, and scale.

4. Honest visual interpretation

Use colorblind-safe palettes, avoid truncated axes when they mislead, show individual points when possible, and state limitations in the caption.

Caption template: Dataset; preprocessing; n; statistic; software/version; interpretation; limitation.

5. Choosing the Right Plot Type

The most common mistake in data visualization is choosing a plot type that obscures the data's structure. Use this decision guide:

6. A Complete Plotting Example in Python

import matplotlib.pyplot as plt
import numpy as np
import pandas as pd

# Create example QC data
np.random.seed(42)
qc_data = pd.DataFrame({
    "sample": [f"S{i:02d}" for i in range(1, 13)],
    "total_reads_M": np.random.normal(25, 5, 12),
    "mapping_rate": np.random.uniform(0.85, 0.98, 12),
    "condition": ["control"]*6 + ["treated"]*6
})

# Create a publication-quality figure
fig, axes = plt.subplots(1, 2, figsize=(12, 5))

# Panel A: Read depth per sample
colors = ["#4C72B0" if c == "control" else "#DD8452" for c in qc_data["condition"]]
axes[0].bar(qc_data["sample"], qc_data["total_reads_M"], color=colors)
axes[0].set_xlabel("Sample")
axes[0].set_ylabel("Total Reads (millions)")
axes[0].set_title("A. Sequencing Depth")
axes[0].tick_params(axis="x", rotation=45)
axes[0].axhline(y=20, color="red", linestyle="--", alpha=0.5, label="QC threshold")
axes[0].legend()

# Panel B: Mapping rate distribution by group
for cond, color in [("control", "#4C72B0"), ("treated", "#DD8452")]:
    subset = qc_data[qc_data["condition"] == cond]
    axes[1].hist(subset["mapping_rate"], bins=6, alpha=0.6, color=color, label=cond)
axes[1].set_xlabel("Mapping Rate")
axes[1].set_ylabel("Frequency")
axes[1].set_title("B. Alignment Quality")
axes[1].legend()

plt.tight_layout()
plt.savefig("results/figures/qc_summary.webp", dpi=300, bbox_inches="tight")
plt.show()

7. Color and Accessibility

Approximately 8% of males and 0.5% of females have some form of color vision deficiency. Use colorblind-safe palettes:

# Colorblind-safe palettes
import matplotlib.pyplot as plt

# Option 1: Use built-in colorblind-safe colormap
plt.set_cmap("viridis")

# Option 2: Seaborn colorblind palette
import seaborn as sns
sns.set_palette("colorblind")

# Option 3: Manually specify accessible colors
accessible_colors = {
    "blue": "#0072B2",
    "orange": "#E69F00",
    "green": "#009E73",
    "vermilion": "#D55E00",
    "sky_blue": "#56B4E9"
}

8. Figure Quality Standards

Practical Exercise

Create one QC plot and one between-group plot using a small table. Add a caption that states n, units, preprocessing, and one limitation.

Pass criteria: A reader can understand the axes, groups, sample count, and what the figure does not prove.

Troubleshooting

If colors are indistinguishable, use a palette designed for color-vision accessibility and add shapes or labels rather than relying on color alone.

Next Steps

Continue with scRNA-seq Basics and Research Reporting and Interpretation. Record the software versions, dataset or example inputs, and any decisions you made.