DESeq2 results columns explained
A DESeq2 results table has six columns per gene. Two of them, the log2 fold change and the adjusted p-value, do most of the work.
The six columns
| Column | What it is |
|---|---|
baseMean | The gene's average normalised read count across all samples, in both conditions. It tells you how highly expressed the gene is, not how much it changed. |
log2FoldChange | The estimated change between conditions on a log2 scale. 1 means doubled, -1 means halved. |
lfcSE | The standard error of that estimate. A large value means the fold change is uncertain. |
stat | The test statistic. For the default Wald test it is the fold change divided by its standard error. |
pvalue | The probability of a change at least this large if the gene were truly unchanged. |
padj | The p-value adjusted for testing thousands of genes at once, using the Benjamini-Hochberg method. Use this one to call genes significant. |
Which direction is "up"?
The fold change compares one condition with a reference. With treated versus control, a positive value means higher in treated. The comparison is printed at the top of the results in R, for example condition treated vs control. If someone else ran the analysis, ask which condition was the reference, because the sign of every fold change depends on it.
Why some padj values are NA
DESeq2 sets padj to NA on purpose in three cases:
- The gene had zero counts in every sample.
- The gene's average count was too low to have a realistic chance of reaching significance. DESeq2 filters these out to improve the power for the remaining genes.
- A sample had an extreme outlier count for that gene.
These genes were not tested in any useful sense, so leave them out of downstream analysis.
Why padj matters more than pvalue
Testing 20,000 genes at a p-value cutoff of 0.05 would flag about 1,000 genes by chance alone. The adjusted p-value controls the false discovery rate instead: at padj below 0.05, about 5% of the genes you call significant are expected to be false positives.
What to do with the table next
The usual next steps are a volcano plot to see the overall picture, then pathway enrichment to find what the changed genes have in common. Keep the full table, including the non-significant genes, because enrichment needs to know every gene that was tested.
Try it on your own results. The DE Results tool on ExpressionBench turns one results table into a volcano plot, a gene table and pathway enrichment. It is free, and your data stays in your browser.
Open the tool