Core promoter grammar breakdown by molecular genetics and deep learning model approaches
Moonshot Award
Hybrid Award (Proposal Development & Innovation Development)
Abstract
Gene transcription and its regulation is critical for an organism’s development and homeostasis. Promoters, a type of transcription regulatory element (TRE), control where and how efficiently transcription begins in human genes.
Our group showed that transcription cycle is regulated mainly at two controlled steps:
- RNA Pol II recruitment & pausing
- release of paused Pol II into productive elongation
Deep learning approaches aided by denoised input can uncover the “sequence grammar” of promoters, linking architecture to regulation of transcription cycle, and disease mechanisms. Our data indicate that promoters isolated from their native loci and separated from other TREs show distinct regulatory behaviors that may be related to the promoter category.
Our workflow enables the generation of comprehensive, noise-free promoter datasets, improving the accuracy of deep learning model predictions. These results support us to secure a NIH R01 (1R01GM166347-01) and a NIH RM1 (2RM1GM139738-06).