Unveiling gene expression histonic regulative patterns by hyperplanes clustering

In targeted cancer therapy, great relevance is assumed by data-driven investigations on the fundamental mechanisms by which epigenetic modifications cooperate to regulate the transcriptional status of genes. At the high resolution level of genome-wide studies, only general, mean regulative motifs are drawn, with possible multi-functional co-regulative roles remaining concealed. In order to retrieve sharper and more reliable regulative patterns, in this work we propose the application of Kplane regression to partition the set of protein coding genes into clusters with shared regulative mechanisms. Completely data-driven, the approach has computed clusters of genes significantly better fitted by specific linear models than by single regression, and characterized by distinct histonic input patterns and mean measured expression values.