Regulatory grammar in human promoters uncovered by MPRA-based deep learning

Area of Science:

• Genomics
• Computational Biology
• Molecular Biology

Background:

• Gene promoters are crucial regulatory elements controlling transcription levels for cellular homeostasis and signal response.
• Accurate prediction of genome-wide gene expression from regulatory element sequences remains a significant challenge in genomics.
• Understanding promoter function is key to deciphering cellular responses and maintaining biological balance.

Purpose of the Study:

• To present the Promoter Activity Regulatory Model (PARM), a novel deep-learning approach for predicting gene expression.
• To develop a cell-type-specific model capable of predicting autonomous promoter activity from DNA sequence alone.
• To enable the design of synthetic strong promoters and systematically analyze transcription factor binding and interactions.

Main Methods:

• Development of PARM, a cell-type-specific deep-learning model trained on Massively Parallel Reporter Assays (MPRAs) of human promoter sequences.
• Utilizing MPRAs to query and analyze the activity of thousands of human promoter sequences.
• Leveraging PARM for systematic identification of transcription factor binding sites and analysis of regulatory interaction rewiring.

Main Results:

• PARM accurately predicts autonomous promoter activity across the genome from DNA sequence, being experimentally and computationally lightweight.
• The model can design novel, strong synthetic promoters.
• Systematic analysis revealed transcription factor positional preferences for activating/repressive functions and complex motif-motif interactions, with detected rewiring upon cellular stimuli.

Conclusions:

• PARM offers an economically viable strategy for a deeper understanding of dynamic human promoter regulation by transcription factors.
• The model facilitates the identification of key regulatory elements and their functional grammar.
• This approach enhances the study of gene regulation in various cellular contexts and responses to stimuli.