Insights into DNA-protein interactions and functional clustering by studying motif distributions in the genome

Abstract

Interacting proteins usually bind 5-6 base pairs of DNA. We looked at distributions of 5/6-mer DNA motifs in whole chromosomes and smaller sections across the whole genome to get insights into how proteins read genomic sequences. The distribution of motifs in the genome is non-random, as established by their observed to expected (OE) ratios at all examined length scales. We correlated the motif distributions in promoter regions of genes to one another and found 5 chromosomes implicated in Robertsonian translocations to be strongly linked. Correlations of OE ratios of motifs in gene promoters gave insights into gene regulation and function. Our analysis identifies common transcription factor binding motifs with high OE ratio values between gene pairs. We also show how one could build a possible network of interactions between genes using the OE ratios. In general, correlating genomic regions by motif distribution comparisons alone is rife with functional information. We extended this analysis in the context of non-random spatial positioning and localisation of the chromosome territories (CTs). The nuclear lamins lining the nuclear envelope potentially interact with the genome, forming high-frequency contact regions termed lamina-associated domains (LADs). These LADs, associated with CTs, exhibit a peripheral organization. Using experimentally available data, we have identified repetitive and evolutionary conserved motifs across LADs in different chromosomes. These are the likely motif candidates responsible for the peripheral or interior tethering of the chromosomal segments in the cell nucleus. Using the motif distributions, we predict LAD and interLAD(iLAD) regions from genomic sequences by designing a Viterbi implementation of the Hidden Markov Model.