# Make a volcano plot from your differential expression results

> Upload the results table from DESeq2, edgeR, limma or Cuffdiff. Yila checks which way the fold change points, applies your thresholds, colours and labels the genes that matter, and saves the counts and the script.

URL: https://yila.ai/playbooks/volcano-plot

## What you get

- **Volcano plot** (PNG + SVG): Fold change against −log10 adjusted p, threshold lines, up and down genes coloured, the most significant genes labelled and the counts in the legend.
- **Labelled genes** (CSV): The genes named on the plot with their fold change, p-value and q-value, ready for the figure legend.
- **Counts** (CSV): How many genes were tested, up, down and excluded, so the numbers in your text match the figure.
- **The script** (Python script): The complete plot in Python (or R if you ask), so new thresholds or a new comparison take one line.

## How it works

1. **Upload the results and name the comparison**: Give Yila the table from your differential expression tool and say which group was compared with which. Raw output from DESeq2, edgeR, limma or Cuffdiff works as it is.
2. **Yila checks direction and edge cases**: It reads which group the tool used as the reference and flips the sign if needed, so up really means up. It also finds infinite fold changes and p-values at the tool's floor before plotting.
3. **Get the plot, the tables and the script**: The plot is rendered, checked for overlapping labels and saved with the gene tables and the script. Ask for other thresholds, your own gene list on the labels or journal colours in the same conversation.

## Made for

- **An RNA-Seq results figure**: The overview figure most transcriptomics papers open their results with.
- **Proteomics and metabolomics**: Any table with an effect size and an adjusted p-value per feature works the same way.
- **Checking a public dataset**: Plot a GEO results table to see whether the published genes stand out before you use them.

## Questions

### What is a volcano plot?

A scatter plot of every gene's fold change (x) against the significance of that change, as −log10 of the p or q-value (y). Genes that change a lot and reliably end up in the upper corners, which is what the plot is for.

### Why does the direction of the fold change matter?

Tools define fold change relative to a reference group, and the reference is not always the one you expect. Cuffdiff, for example, reports sample 2 over sample 1. In the example the table was untreated over dexamethasone, so plotted as it was, every induced gene would have appeared as repressed. Yila checks this and tells you what it did.

### Which thresholds should I use?

A common choice is |log2 fold change| ≥ 1 (a two-fold change) with an adjusted p-value below 0.05, which is what the example uses. Your field may prefer other values; say so in the request.

### What happens to genes with infinite fold change or p = 0?

Infinite fold changes come from genes not detected in one group. They are excluded from the plot and counted, rather than stacked on the axis edge. p-values at the software's floor are drawn at that floor and labelled as such.

### Can I choose which genes are labelled?

Yes. By default the most significant are labelled; you can give a list of genes of interest, or ask for the top genes by fold change instead.

### R or Python?

Either. The example used Python with matplotlib; in R, Yila uses ggplot2 with ggrepel or EnhancedVolcano.

### Where does the example data come from?

From the public GEO series GSE52778, the RNA-Seq data behind Himes et al., PLOS ONE 2014. Yila never uses users' data as examples.

---

- [Make a volcano plot from your differential expression results](https://yila.ai/playbooks/volcano-plot)
- [llms.txt](https://yila.ai/llms.txt)
