TIGP (BIO)—Understanding the genome-wide transcription cis-regulation in Drosophila using deep learning

Abstract

Transcription regulation in metazoans relies on the intricate interactions between transcription factors (TFs) and modular DNA sequences known as cis-regulatory modules (CRMs). These CRMs, encompassing enhancers, promoters, and insulators, orchestrate the gene expression dynamics essential for cellular functions. However, traditional experimental approaches for CRM study are limited in throughput and cost-effectiveness, necessitating computational methods to help carry out genome-wide CRM research. Existing in silico tools often focus on specific CRM types, leading to errors when considering the multifaceted roles of CRMs. To address this, we designed deep learning approaches to advance CRM annotation and interaction prediction. We first constructed the regCNN model that leverages local patterns in TF binding motifs and epigenetic profiles for CRM identification. Then, for the identified CRMs, a tool called CRM Function Annotator (CFA) integrating epigenetic profiling was developed to label their transcriptional roles. Additionally, we introduce a deep learning-based method for identifying CRM interactions, demonstrating its effectiveness in recognizing various CRM interactions beyond enhancer-promoter interactions. Since understanding the target genes of CRMs is challenging for enhancers and insulators, we designed a deep learning-based pipeline that comprehensively identifies all genres of CRM-target-gene pairs, including promoter targets, enhancer targets, and insulator targets. Lastly, since current CRM experimental results are fragmentary across literature, to facilitate text mining for CRM-related information, we introduce DMLS (Drosophila Modular transcription Literature Screener). This text-mining tool extracts target genes and regulatory TFs associated with CRMs from Drosophila transcription regulation-related articles. Altogether, the designed computational frameworks provide valuable resources for dissecting metazoan transcriptional regulatory networks, providing hope for the diagnosis of genetic disorders and advancing our understanding of cellular transcription dynamics.

附件下載