Item

Abstract

A deep understanding of the dynamics of gene regulatory net- works relies on the availability of models faithfully predicting the binding of transcription factors (TFs) to their DNA targets. Posi- tion Weight Matrices (PWM) describe nucleotide preference at each position of a TF binding site (TFBS) but, in addition to site presence, in vivo binding is modulated by several factors (TF binding competition, DNA accessibility, etc). In any case, an essential prerequisite is a good PWM that correctly establishes the interaction with potential TFBS. ChIP-chip/Seq experiments give a very useful TF binding picture but obtaining these data for all TFs, each tissue, developmental process, condition and developmental stage is unapproachable, even more in non model organisms. Optimization of PWMs allows the identification of gene regulatory networks in non-model species, and the possibil- ity of manipulation of agronomically important traits. Several algorithms are available to establish PWMs but they render significantly different results, so finding an evaluation method for the different models is prioritary. We propose a method, SIMCHIP, that simulates the binding for a given TF based on ChIP-chip/seq or SELEX data, and compares the out- put with in vivo data to evaluate PWM performance peak by peak. As a proof of concept, we have used this method on the plant-specific TF LEAFY and on the human TF STAT1. SIMCHIP and other binding analysis tools based on PWMs are available at http://biodev.cea.fr/morpheus.