Wavelet-based method to disentangle transcription- and replication-associated strand asymmetries in mammalian genomes

During genome evolution, the two strands of the DNA double helix are not subjected to the same mutation patterns. This mutation bias is considered as a by-product of replicative and transcriptional activities. In this paper, we develop a wavelet-based methodology to analyze the DNA strand asymmetry profiles with the specific goal to extract the contributions associated with replication and transcription respectively. In a first step, we use an adapted N-shaped analyzing wavelet to perform a multi-scale pattern recognition analysis of the sum of the TA and GC skews along human chromosomes. This method provides an objective segmentation of the human genome in skew domains of ≃1 Mbp characteristic size, bordered by two putative replication origins recognized as large amplitude upward jumps in the noisy skew profile. In a second step, we use a least-square fitting procedure to disentangle, in these skew domains, the small-scale (the mean human gene size ≃30 kbp) square-like transcription component from the global N-shaped component induced by replication. When applying this procedure to the 22 human autosomes, we delineate 678 replication domains of mean length spanning 33.8% of the human genome and we predict 1062 replication origins. When investigating the distribution of transcription-associated skew inside the replication N-domains, we reveal some dependence upon the distance to the putative replication origins located at N-domain extremities, the closer the genes to the origin, the larger their transcription bias as the signature of a higher transcriptional activity in the germ-line. As a comparative analysis, we further apply our wavelet-based methodology to skew profiles along the mouse chromosomes. The striking similarity of the results in human and mouse indicates that the remarkable gene organization observed inside the human replication N-domains is likely to be a general feature of mammalian genomes.