High-recombining genomic regions affect demography inference

Inference of population history of non-model species is important in evolutionary and conser- vation biology. Multiple methods of population genomics, including those to infer population history, are based on the ancestral recombination graph (ARG). These methods use observed mutations to model local genealogies changing along chromosomes. Breakpoints at which genealogies change effectively represent the positions of historical recombination events. How- ever, inference of underlying genealogies is difficult in regions with high recombination rate relative to mutation rate. This is because genealogies cover genomic intervals that are too short to accommodate sufficiently many mutations informative of the structure of the un- derlying genealogies. Despite the prevalence of high-recombining genomic regions in some non-model organisms, such as birds, its effect on ARG-based demography inference has not been well studied. Here, we use population genomics simulations to investigate the impact of high-recombining regions on ARG-based demography inference. We demonstrate that inference of effective population size and the time of population split events is systematically affected when high-recombining regions cover wide breadths of the chromosomes. We also show that excluding high-recombining genomic regions can practically mitigate this effect. Finally, we confirm the relevance of our findings in empirical analysis by contrasting demography inferences applied for a bird species, the Eurasian blackcap ( Sylvia atricapilla ), using different parts of the genome with high and low recombination rates. Our results suggest that demography inference using ARG-based methods should be carried out with caution when applied in species whose reference genomes contain long stretches of high-recombining regions.