MOTIVATION: Accurately inferring transcription factor (TF) activity from single-cell RNA sequencing (scRNA-seq) data remains a fundamental challenge in computational biology. While existing methods rely on statistical models, motif enrichment, or prior-based inference, they often depend on deterministic assumptions about regulatory relationships and rely on static regulatory databases. Few approaches effectively integrate prior biological knowledge with data-driven inference to capture novel, dynamic, and context-specific regulatory interactions. RESULTS: To address these limitations, we develop scRegulate, a generative deep-learning framework leveraging variational inference to estimate TF activities guided by experimental TF-target gene relationships and progressively adapted based on the input scRNA-seq data. By integrating structured biological constraints with a probabilistic latent space model, scRegulate offers a scalable and biologically grounded estimation of TF activity and gene regulatory network (GRN). Comprehensively benchmarking on public experimental and synthetic datasets demonstrates scRegulate's superior ability. Further, scRegulate accurately recapitulates experimentally validated TF knockdown effects on a Perturb-seq dataset for key TFs. Applied to experimental human PBMC scRNA-seq data, scRegulate infers cell-type-specific GRNs and identifies differentially active TFs aligned with known regulatory pathways. scRegulate's TF activity representations capture transcriptional heterogeneity, enabling accurate clustering of cell types. scRegulate is highly efficient, frequently an order of magnitude faster than common baselines. Collectively, our results establish scRegulate as a powerful, interpretable, and scalable framework for inferring TF activities and GRNs from single-cell transcriptomics. AVAILABILITY AND IMPLEMENTATION: Results and scripts available at github.com/YDaiLab/scRegulate.

BACKGROUND: The heterogeneity of cells across tissue types represents a major challenge for studying biological mechanisms as well as for therapeutic targeting of distinct tissues. Computational prediction of tissue-specific gene regulatory networks may provide important insights into the mechanisms underlying the cellular heterogeneity of cells in distinct organs and tissues. RESULTS: Using three pathway analysis techniques, gene set enrichment analysis (GSEA), parametric analysis of gene set enrichment (PGSEA), alongside our novel model (HeteroPath), which assesses heterogeneously upregulated and downregulated genes within the context of pathways, we generated distinct tissue-specific gene regulatory networks. We analyzed gene expression data derived from freshly isolated heart, brain, and lung endothelial cells and populations of neurons in the hippocampus, cingulate cortex, and amygdala. In both datasets, we found that HeteroPath segregated the distinct cellular populations by identifying regulatory pathways that were not identified by GSEA or PGSEA. Using simulated datasets, HeteroPath demonstrated robustness that was comparable to what was seen using existing gene set enrichment methods. Furthermore, we generated tissue-specific gene regulatory networks involved in vascular heterogeneity and neuronal heterogeneity by performing motif enrichment of the heterogeneous genes identified by HeteroPath and linking the enriched motifs to regulatory transcription factors in the ENCODE database. CONCLUSIONS: HeteroPath assesses contextual bidirectional gene expression within pathways and thus allows for transcriptomic assessment of cellular heterogeneity. Unraveling tissue-specific heterogeneity of gene expression can lead to a better understanding of the molecular underpinnings of tissue-specific phenotypes.