IIT Home Page CNR Home Page

CMStalker: a combinatorial tool for composite motif discovery

Controlling the differential expression of many thousands different genes at any given time is a fundamental task of metazoan organisms and this complex orchestration is controlled by the so-called regulatory genome encoding complex regulatory networks: several Transcription Factors bind to precise DNA regions, so to perform in a cooperative manner a specific regulation task for nearby genes. The in silico prediction of these binding sites is still an open problem, notwithstanding continuous progress and activity in the last two decades. In this paper we describe a new efficient combinatorial approach to the problem of detecting sets of cooperating binding sites in promoter sequences, given in input a database of Transcription Factor Binding Sites encoded as Position Weight Matrices. We present CMStalker, a software tool for composite motif discovery which embodies a new approach that combines a constraint satisfaction formulation with a parameter relaxation technique to explore efficiently the space of possible solutions. Extensive experiments with twelve data sets and eleven state-of-the-art tools are reported, showing an average value of the correlation coefficient of 0.54 (against a value 0.41 of the closest competitor). This improvements in output quality due to CMStalker is statistically significant.

IEEE/ACM Transactions on Computational Biology and Bioinformatics, 2015

Autori esterni: Karina Tillan (Universita' di Modena e Reggio Emilia)
Autori IIT:

Manuela Montangero

Foto di Manuela Montangero

Tipo: Articoli su riviste ISI
Area di disciplina: Computer Science & Engineering

Attività: Biologia computazionale