About PyDESeq2

DESeq2 [LHA14] is the reference method for differential expression analysis of bulk RNA-seq data. It models raw counts with a negative binomial distribution, shares information across genes to stabilise the per-gene dispersion estimates, and tests log fold-changes with a Wald test. It is written in R.

PyDESeq2 [MTCA23] reimplements that method in Python, on AnnData objects, so that a differential expression step fits inside a Python analysis without leaving the language.

Scope

Current features broadly correspond to the default settings of DESeq2 (v1.34.0):

• single-factor and multi-factor analysis, with categorical or continuous factors

• Wald tests, with Cooks filtering and independent filtering

• optional apeGLM log fold-change shrinkage [ZIL19]

• variance-stabilising transformation

Differences from DESeq2

PyDESeq2 is a reimplementation from scratch rather than a port, so retrieved values may differ slightly from those of DESeq2, and not every DESeq2 feature exists. If a feature you rely on is missing, open an issue on GitHub.