Regulatory and Network Analysis

 

Regulatory and Network Analysis

From expression to regulation

Identifying differentially expressed genes is a starting point for transcriptomic and proteomic experiments. The next question is why those genes are differentially expressed. Which transcription factors are driving the changes, and what are the regulatory relationships among them? Gene regulatory network inference attempts to reconstruct those relationships from transcriptomic data, producing a causal model of how gene expression is controlled rather than a list of correlations.

Single-cell GRN inference

pySCENIC (the current Python implementation of SCENIC) infers gene regulatory networks at single-cell resolution by identifying co-expressed transcription factor and target gene modules, called regulons, and scoring each cell for regulon activity. SCENIC+ extends this to paired RNA and ATAC data, using chromatin accessibility to validate that identified binding sites are actually open in the cells where a regulon is active. This substantially improves specificity, particularly in heterogeneous tissues where the same transcription factor may regulate different targets in different cell types.

GRaNIE is an alternative approach designed for tissue-level GRN inference, incorporating eQTL information and sample-level variation to identify regulatory programs that are active across biological replicates rather than in individual cells. It is particularly useful for cohort-level analyses where replicate structure is important.

It is important to note that GRN inference from transcriptomics alone is correlational. Validation of key regulatory relationships requires orthogonal data (e.g., ChIP-seq, CUT&RUN, or perturbation experiments). 3DG reports inferred networks with appropriate uncertainty quantification and flags the most biologically plausible regulatory relationships for experimental follow-up.

Spatial regulatory program mapping

When combined with spatial transcriptomics, regulon activity scores can be projected onto tissue sections to map where specific regulatory programs are active. This is informative in tissues where cell identity is partially defined by location, such as cortical layers in the brain, skin compartments from epidermis to dermis, or tumor regions with distinct microenvironmental niches.

Spatially-resolved regulatory analysis connects the molecular logic of gene regulation to the anatomical organization of the tissue, identifying microenvironments where specific transcription factors are active and linking those regions to cellular identities and functional states.