Pathway Activity Analysis Pipeline

This page documents the gene set activity analysis pipeline. See Methods for the manuscript prose.

Method

• Framework: Univariate Linear Model (ULM) in the decoupler R package
• Input: Wald z-scores from DESeq2 (log₂FC / SE)
• Pre-filtering: None — all genes with non-missing statistics included to preserve full null distribution

Gene Sets

• Source: MSigDB v2026.1 via OmniPath (downloaded January 2026)
• Collections: Hallmark, KEGG, Reactome, WikiPathways, BioCarta, Gene Ontology Biological Process
• Size filter: 10–400 overlapping genes per gene set
• Total tested: 7,774 gene sets

Model

ULM fits a linear model for each gene set, regressing gene-level statistics against gene-set membership to estimate pathway activity scores.

Significance

• Model-derived p-values corrected via Benjamini–Hochberg FDR
• Significance threshold: FDR < 0.05

Exclusions

• Dendritic cell populations excluded due to insufficient cells/pseudobulk profiles

Related Pages

• Methods — Manuscript Draft
• DEG Pipeline
• Regulatory Network Inference