What log2FoldChange Means in DESeq2 Output
A plain explanation of the log2FoldChange column in DESeq2 results, how the shrunken estimate differs from a ratio of means, and how to compute and check it yourself.
12 posts
A plain explanation of the log2FoldChange column in DESeq2 results, how the shrunken estimate differs from a ratio of means, and how to compute and check it yourself.
What the axes mean, why you should shrink fold changes before plotting, and how to diagnose the odd shapes people run into with plotMA.
An MA plot puts fold change against mean expression; a volcano plot puts fold change against significance. Both are scatter plots of the same differential expression table, and each catches a different failure mode.
A log2 fold change of 1 means the measured quantity doubled between conditions. This page explains how that number is computed, why it is shrunk, and how to decide whether a doubling is real.
Log2 fold change is log2(treated / control). A log2FC of 4 is a 16-fold increase. This page covers the formula, the conversions, the code, and the failure modes that make a naive calculator wrong.
log2FC is the base-2 logarithm of a ratio between two expression estimates. This page explains the arithmetic, the shrinkage that DESeq2 and limma apply, and the failure modes that make a large log2FC meaningless.
A step-by-step guide to differential expression analysis with DESeq2, from salmon quantification through shrunken log fold changes, written for someone analyzing their own RNA-seq.
baseMean is the average of normalized counts across all samples. Thresholding on it after results() is the wrong move; filter raw counts before fitting and let independent filtering do the rest.
Web platforms are good at the downstream half of RNA-seq (differential expression, enrichment, plots) and bad at the upstream half (FASTQ to counts). Here is the split we'd use, with the commands, parameters, and failure modes.
A step-by-step guide to producing a volcano plot from a count matrix with DESeq2 and ggplot2, choosing thresholds that mean something, and reading the resulting shape correctly, including the single-cell and mass-spectrometry variants.
A practical walkthrough of differential expression analysis in R, from transcript quantification through DESeq2 and limma models, with the design decisions and failure modes that determine whether the gene list means anything.
A post hoc |log2FC| > 1 filter is the wrong tool. Test against a fold-change threshold directly with DESeq2's lfcThreshold or edgeR's glmTreat, shrink effect sizes before ranking, and pick the threshold from your own replicate structure.