If you’ve done any kind of differential analysis — RNA-seq, proteomics, CRISPR screens — you’ve stared at a volcano plot. Done well, they make thousands of features readable at a glance.
What’s on a volcano plot
A volcano plot is a scatter plot with two axes:
- X-axis: Effect size — typically log2 fold-change. Negative values mean the feature is decreased in the treatment; positive values mean increased.
- Y-axis: Statistical significance — typically -log10(p-value) or -log10(adjusted p-value). Higher = more significant.
Each dot is one gene, protein, or feature. The “volcano” shape comes from many features clustering at low fold-change with low significance, and a few rising up at the top corners.
The four regions
| Region | Interpretation |
|---|---|
| Upper right | Increased, significant — likely up-regulated |
| Upper left | Decreased, significant — likely down-regulated |
| Lower middle | Small effect, not significant — noise/no change |
| Lower left/right | Large effect but not significant — usually low-confidence |
Setting thresholds
- Fold-change cutoff: Typically |log2FC| > 1 (2-fold change), but biology-dependent
- Significance cutoff: Adjusted p-value < 0.05 (or < 0.01)
Always use adjusted p-values (FDR or BH correction), not raw p-values, in genome-scale experiments. With 20,000 genes tested, you expect 1,000 false positives at p < 0.05 by chance alone.
Common misinterpretations
- “Highly significant” doesn’t mean “biologically important.” A gene with p = 10⁻²⁰⁰ and fold-change of 1.05 may be very real but irrelevant.
- “Large fold-change” doesn’t mean “real.” A gene with log2FC = 6 and p = 0.5 is probably noise.
- The strongest hits combine effect size and significance — they sit in the upper corners.
Improving readability
- Label only key genes — top hits, candidates of interest, outliers
- Color by category (up, down, not significant) or by gene set
- Use transparency to show density without obscuring sparse outliers
- Adjust axis limits — cap p-values at, say, 10⁻³⁰⁰ for plotting
- Annotate gene counts in each region
Tools for volcano plots
- R: EnhancedVolcano (Bioconductor) — feature-rich and publication-ready
- R / ggplot2: for full custom control
- Python: matplotlib, seaborn, or bioinfokit
- GraphPad Prism: straightforward for smaller datasets
What to do after a volcano plot
- Pathway analysis on the significant gene list (GSEA, Enrichr, Reactome)
- Validation by qPCR or western blot for top hits
- Functional follow-up on candidates of interest
- Comparison to public datasets for replication
A volcano plot summarizes thousands of statistical tests in one image. The genes that matter are usually those with both meaningful effect sizes and strong significance — the upper corners.
