PanSel: a tool finding conserved/divergent regions

Overview & Synopsis

Introduction Although pangenomic studies have been published for almost two decades in prokaryotes, they are relatively new for multi-cellular eukaryotes, and changing the framework from a genome supplemented with variations to a variation graph is no simple task. Notably, it requires adapting all the tools which were developed for the previous framework, and they are numerous. Studying conservation of genomic sequence, i.e. sequences that evolve slower (or faster) than the expected natural drift, is one of them. Methods, made for linear genomes, include phastCons [4], GERP and GERP++ [1], and phyloP [3]. These tools are heavily used when one wants to check whether a given gene is subject to selection. They rely on two types of information: a multiple sequence alignment, and a phylogeny. In PanSel, we adapt the idea of the quantification of the genomic conservation to variation graphs, using simple ideas. First, we use sliding windows of fixed size on the reference haplotype. Second, we estimate the average number of edit distance between each haplotype on each window. Last, significantly under- and over-conserved windows are detected by fitting a mixture model. Results PanSel reads a GFA file representing one chromosome. Given a window size s, it tries to detect conserved segments (named name boundary segments hereafter) distant by s nucleotides on the reference path. For each pair of boundary segments, it extracts the sub-path for each haplotype. It estimates the average edit distance between each pair of sub-paths. Considering two paths, we compute the Jaccard index of the nodes in the sub-paths. The distribution of the edit distance is then fitted with a simple mixture model. Significantly constrained and divergent regions are then computed. We compared our results on the Human pangenome [2], with available annotations: conservation, coding exons, predicted chromatin states. We show that our tool confirms expected results: genes are more constrained that non-coding regions. More interestingly, we found that, conserved regions are related to very general molecular functions, biological processes or cellular compartment. Divergent regions are related to more specific molecular functions, biological processes or cellular compartment, which are related to immunity. PanSel, written in C++11 with no dependency, is available at https://github.com/mzytnicki/pansel. References [1] Eugene V. Davydov et al. “Identifying a High Fraction of the Human Genome to be under Selective Constraint Using GERP++”. In: PLoS Computational Biology 6.12 (2010), e1001025. [2] Wen-Wei Liao et al. “A draft human pangenome reference”. In: Nature 617.7960 (2023), pp. 312–324. [3] Katherine S. Pollard et al. “Detection of nonneutral substitution rates on mammalian phylogenies”. In: Genome Research 20.1 (2009), pp. 110–121. [4] Adam Siepel et al. “Evolutionarily conserved elements in vertebrate, insect, worm, and yeast genomes”. In: Genome Research 15.8 (2005), pp. 1034–1050.