Global inference of DNA-protein binding motifs reversible Jump Hamiltonian Monte Carlo

Abstract

Listen

Translate article

Transcription factor proteins locate their genetic targets by binding DNA at sequence specific motifs, creating binding footprints distinct for each protein’s binding. However, inferring motifs from such footprints is not trivial, particularly for datasets where an unknown number of motifs may be active simultaneously. Such a situation is observed, for example, in ChIP-seq data for promiscuous proteins or global protein binding data extracted from ATAC-seq or IPOD experiments. We have designed an algorithm based on Reversible Jump Hamiltonian Monte Carlo (RJHMC) to enable simultaneous inference of the number and identity of motifs acting to generate an occupancy profile.