pydeseq2

Bulk RNA-seq differential expression converts a non-negative integer samples-by-genes counts DataFrame and matching metadata into DE gene calls. You provide a Wilkinson-style design string and contrast specification to compute Wald statistics and FDR-adjusted p-values.

Yes, multi-factor designs and batch covariates are supported by specifying the appropriate Wilkinson-style design formula. This allows the model to account for confounding variables while testing your specified contrast for differential expression.

Optional LFC shrinkage provides more stable effect sizes for bulk RNA-seq visualization and interpretation. Applying this during differential expression testing reduces noise in log2 fold change estimates for downstream diagnostic plots like volcano or MA plots.

Yes, differential expression testing requires counts as non-negative integer values in a samples-by-genes DataFrame. Normalized or non-integer inputs are not valid for computing accurate Wald statistics and FDR-adjusted p-values.

A valid contrast specification is required for testing differential expression even in single-factor designs. You must ensure the contrast matches a variable present in your Wilkinson-style design formula and metadata to successfully compute gene calls.

This workflow is limited to bulk RNA-seq data and requires matching sample indices between counts and metadata. It does not process single-cell RNA-seq or accept non-integer count matrices for computing Wald statistics and FDR-adjusted p-values.