The distribution of branch duration and detection of inversions in ancestral recombination graphs

Abstract

AbstractRecent breakthroughs in the reconstruction of ancestral recombination graphs (ARGs) have meant that inference of genome-wide genealogies from large sequencing datasets is now possible. This development has also opened several new directions where research is urgently needed: improving our understanding of which aspects of genealogies are (and which are not) faithfully captured by reconstruction tools, as well as the development of inference methods that directly use ARGs as inputs. By deriving the distribution of branch duration under the SMC’ model, defined as the genomic interval spanned by a branch of the ARG until it is disrupted by recombination, we benchmark the quality of ARGs produced by several reconstruction tools: ARGweaver, Relate, tsinfer and tsdate, and ARG-Needle. Further, we develop an accurate and powerful ARG-based test for the presence of chromosomal inversions, based on the idea that suppression of recombination in individuals heterozygous for an inversion has a detectable impact on branch duration.