decoupler-tf-activity

Transcription factor activity is inferred from bulk RNA-seq differential expression by applying a univariate linear model to a matrix of t-statistics using CollecTRI regulatory network priors. This process generates per-TF activity scores and p-values across multiple contrasts.

The CollecTRI network provides the regulatory weights required for TF activity analysis. It acts as a prior knowledge base fetched via OmnipathR, linking transcription factors to their target genes to model regulatory relationships during the univariate linear model calculation.

DoRothEA can be used as an alternative regulon source instead of CollecTRI. The workflow provides options to switch between these regulatory networks and includes species mapping capabilities to handle different organism data during the TF activity inference.

Calculating TF activity scores across multiple contrasts requires a numeric matrix of t-statistics with genes as rows and contrasts as columns. The decoupleR ULM workflow processes this matrix with minsize filtering to exclude small regulons, returning a master table of comparative scores and p-values.

TF activity inference excludes transcription factors with small regulons through minsize filtering. This precautionary step removes TFs with insufficient target gene counts in the CollecTRI network to ensure statistical reliability of the univariate linear model activity scores.