Analysis of short- and long-range interactions within potential binding sites notably extends the fraction of verified peaks in ChIP-seq data

Poster (download)

299

Anton Tsukanov¹, Victor Levitsky², Tatyana Merkulova^3
1ICG SB RAS, Novosibirsk, Russia, tsukanov@bionet.nsc.ru
²ICG SB RAS, Novosibirsk, Russia, levitsky@bionet.nsc.ru
³ICG SB RAS, Novosibirsk, Russia, merkulova@bionet.nsc.ru

We developed pipeline for integrative application of various de novo motif search tools to massive sequencing data. While traditional position weight matrices (PWMs) neglect dependencies between positions of motifs, the short-range interactionsв markov models BAMM/InMode permit only local dependencies, the long-range interaction model SiteGA allows dependencies between arbitrary positions. The massive analysis of ChIP-seq data have shown that the models BAMM/InMode and PWMs recognized similar and significantly overlapped peaks; but the notable fractions of peaks predicted by SiteGA model were not predicted by other models.