QPS -- quadratic programming sampler, a motif finder using biophysical modeling

Abstract

We present a Markov chain Monte Carlo algorithm for local alignments of nucleotide sequences aiming to infer putative transcription factor binding sites, referred to as the quadratic programming sampler. The new motif finder incorporates detailed biophysical modeling of the transcription factor binding site recognition which arises an intrinsic threshold discriminating putative binding sites from other/background sequences. We validate the principal functioning of the algorithm on a sample of four promoter regions from Escherichia coli. The resulting description of the motif can be readily evaluated on the whole genome to identify new putative binding sites.