TRAPT
Zhang G, Song C, Yin M, Liu L, Zhang Y, Li Y, Zhang J, Guo M, Li C
2025 · Nature communications
Epigenomics-first deep learning maps a query gene set to ranked upstream TR activity using large ChIP/ATAC/H3K27ac compendia.
Abstract
From the original paper, Nature communications · PubMed
It is challenging to identify regulatory transcriptional regulators (TRs), which control gene expression via regulatory elements and epigenomic signals, in context-specific studies on the onset and progression of diseases. The use of large-scale multi-omics epigenomic data enables the representation of the complex epigenomic patterns of control of the regulatory elements and the regulators. Herein, we propose Transcription Regulator Activity Prediction Tool (TRAPT), a multi-modality deep learning framework, which infers regulator activity by learning and integrating the regulatory potentials of target gene cis-regulatory elements and genome-wide binding sites. The results of experiments on 570 TR-related datasets show that TRAPT outperformed state-of-the-art methods in predicting the TRs, especially in terms of forecasting transcription co-factors and chromatin regulators. Moreover, we successfully identify key TRs associated with diseases, genetic variations, cell-fate decisions, and tissues. Our method provides an innovative perspective on identifying TRs by using epigenomic data.
Summary
Level-1 epigenomic TR-prioritization: query gene set → TR-RP/Epi-RP → D-RP (CVAE/VGAE) and U-RP (distillation + sparse-group lasso) → I-RP/AUC ranked TR activities for TFs, co-factors, and chromatin regulators; validated on KnockTF, KnockTF, Lisa/BART/i-cisTarget/ChEA3 TR-recovery benchmark, and disease/cell-fate case studies; no head-to-head benchmark against other cell-reprogramming methods.
Why this class
Level 1 because the method scores candidate regulators through static regulatory-network / epigenomic influence on a query gene set, without computing the post-intervention state . Cited in §3.2 of the review alongside ANANSE-class tools.