Institute of Information Science, Academia Sinica

Quantification of selective pressures on regulatory sequences is a central question in studying the evolution of gene regulatory networks.  Previous methods focus primarily on single sites rather than motif sequences.  We propose a method of evaluating the strength of natural selection of a motif from a family of aligned sequences.  The method is based on a Poisson process model of neutral sequence substitutions and derives a birth-death process of the motif occurrence frequencies.  The selection coefficient is treated as a penalty for the motif death rate.  We demonstrate that the birth-death model closely approximates statistics generated from simulated data and the Poisson process assumption holds in mammalian promoter sequences.  Furthermore, we show that a considerably higher portion of known transcription factor binding motifs possess high selection coefficients compared to negative controls with high occurrence frequencies on promoters.  Preliminary analysis supports the potential applications of the model to identify regulatory sequences under selection.