Modelling gene regulation networks via multivariate adaptive splines.

After the completion of sequencing for dozens of genomes, as well as the draft of human genome, a major challenge is to characterize genome-wide transcriptional regulation networks. Identification of regulatory functions for transcription factor binding sites in eukaryotes greatly enhances our understanding of the networks, as it has been done extensively under various physiological conditions in yeast. We propose a novel approach based on multivariate adaptive splines to modelling regulatory roles of motifs in gene expression time series data. By applying the proposed approach on two meiotic datasets, we identified well-documented motifs as well as some novel putative motifs that are involved in the transcriptome reprogramming. In addition to identifying single regulatory motifs, we also modelled and unravelled motifs that manifest their effects through coupling with others in regulatory networks. Our findings reveal the potential of multivariate adaptive splines in deciphering complex and important transcriptional regulatory networks in eukaryotes.