How to read a volcano plot

Resources · Tutorial · 6 min read

A volcano plot is the single most common figure in RNA-seq differential expression — and one of the most misread. Here is exactly what it shows and how to read it with confidence.

What a volcano plot is

A volcano plot is a scatter plot that summarizes a differential expression test across every gene at once. Each dot is one gene, placing two numbers on the same picture:

• x-axis — log₂ fold change: how much the gene’s expression changed between your two conditions. Zero means no change; positive is up in the test group, negative is down.
• y-axis — −log₁₀ of the p-value: how strong the evidence is that the change is real. Higher up the plot means a smaller p-value.

The name comes from the shape: a cloud of non-significant genes sits near the bottom center, and significant genes erupt up and out to the sides — like a volcano.

Why the axes are logged

Both axes are on a log scale for a reason. A log₂ fold change of 1 is a doubling, 2 a four-fold increase, −1 a halving — so up- and down-regulation are symmetric around zero. And −log₁₀ turns tiny p-values (0.001, 0.000001) into large, readable heights, so the most significant genes rise to the top.

The thresholds that define “a hit”

Two cutoffs turn the cloud into a shortlist:

• A vertical pair of lines marks a fold-change threshold (often |log₂FC| ≥ 1, a two-fold change). Genes between the lines changed too little to care about.
• A horizontal line marks a significance threshold (often adjusted p < 0.05). Genes below it aren’t statistically convincing.

Only genes in the upper-left (down-regulated, significant) and upper-right (up-regulated, significant) corners clear both bars. Those are your differentially expressed genes.

p-value vs. adjusted p-value

This is the step most people skip. An RNA-seq experiment tests tens of thousands of genes at once, so some will look significant by chance alone. The adjusted p-value (the FDR or q-value, usually from the Benjamini–Hochberg method) corrects for that multiple testing. Always threshold on the adjusted p-value, not the raw one — a good volcano plot puts the adjusted value on the y-axis.

Four mistakes to avoid

• Chasing big fold changes that aren’t significant. A gene far to the right but low down has a large change with weak evidence — often a low-count gene with noisy estimates.
• Trusting tiny changes that are “significant.” With enough replicates, a biologically trivial change can pass a p-value cutoff. Keep a fold-change threshold too.
• Reading raw p-values. Without multiple-testing correction, your hit list is inflated.
• Forgetting direction. Left and right are not interchangeable — check which condition is the baseline so up and down mean what you think.

How to actually use it

Read a volcano plot in three passes: first glance at the overall shape — is the response broad or focused, balanced or skewed? Then adjust the thresholds down to a shortlist you can act on. Finally, label and inspect the top genes and carry them into pathway and enrichment analysis to see the biology.

← Back to Resources