Epiregulon: Inference of single-cell transcription factor activity to dissect mechanisms of lineage plasticity and drug response

Transcription factors (TFs) and transcriptional coregulators represent an emerging and exciting class of targets in oncology. Gene regulatory networks (GRNs) evaluate pharmacological agents against these regulators through quantifying target gene modulation. However, methods relying solely on gene expression cannot adequately model post-transcriptional interference of TF functions. We introduce Epiregulon, a method to predict TF activity at the single-cell level using GRNs constructed from RNA-seq and ATAC-seq data. Considering co-occurrence of gene expression and chromatin accessibility addresses the decoupling of TF activity from gene expression. The incorporation of ChIP-seq data extends inference to transcriptional coregulators lacking defined motifs and facilitates the identification of interaction partners. We uncover divergent cell fate transitions of prostate cancer cells driven by NKX2-1 and GATA6 overexpression. Epiregulon outperforms other methods in its consistency and accuracy in predicting the effects of AR inhibition. Finally, degrader-treated networks recapitulate the context-dependent activity of SMARCA4 consistent with the known etiologies of prostate cancers. By mapping out gene regulation across a multitude of perturbations, Epiregulon can accelerate the discovery of therapeutics targeting transcriptional regulators. ### Competing Interest Statement Aaron Lun, Diana Wu, Shushan Toneyan, Liang Wang, Kerstin Seidel, Jenille Tan, Shang-Yang Chen, Timothy Keyes, Yu Guo, Ciara Metcalfe, Marc Hafner, Christian W. Siebel, Robert Yauch, Shiqi Xie and Xiaosai Yao are or were employees of Genentech Inc.