Genome-Wide Inference of Ancestral Recombination Graphs

The complex correlation structure of a collection of orthologous DNA sequences is uniquely captured by the “ancestral recombination graph” (ARG), a complete record of coalescence and recombination events in the history of the sample. However, existing methods for ARG inference are computationally intensive, highly approximate, or limited to small numbers of sequences, and, as a consequence, explicit ARG inference is rarely used in applied population genomics. Here, we introduce a new algorithm for ARG inference that is efficient enough to apply to dozens of complete mammalian genomes. The key idea of our approach is to sample an ARG of chromosomes conditional on an ARG of chromosomes, an operation we call “threading.” Using techniques based on hidden Markov models, we can perform this threading operation exactly, up to the assumptions of the sequentially Markov coalescent and a discretization of time. An extension allows for threading of subtrees instead of individual sequences. Repeated application of these threading operations results in highly efficient Markov chain Monte Carlo samplers for ARGs. We have implemented these methods in a computer program called ARGweaver. Experiments with simulated data indicate that ARGweaver converges rapidly to the posterior distribution over ARGs and is effective in recovering various features of the ARG for dozens of sequences generated under realistic parameters for human populations. In applications of ARGweaver to 54 human genome sequences from Complete Genomics, we find clear signatures of natural selection, including regions of unusually ancient ancestry associated with balancing selection and reductions in allele age in sites under directional selection. The patterns we observe near protein-coding genes are consistent with a primary influence from background selection rather than hitchhiking, although we cannot rule out a contribution from recurrent selective sweeps.Author Summary: The unusual and complex correlation structure of population samples of genetic sequences presents a fundamental statistical challenge that pervades nearly all areas of population genetics. Historical recombination events produce an intricate network of intertwined genealogies, which impedes demography inference, the detection of natural selection, association mapping, and other applications. It is possible to capture these complex relationships using a representation called the ancestral recombination graph (ARG), which provides a complete description of coalescence and recombination events in the history of the sample. However, previous methods for ARG inference have not been adequately fast and accurate for practical use with large-scale genomic sequence data. In this article, we introduce a new algorithm for ARG inference that has vastly improved scaling properties. Our algorithm is implemented in a computer program called ARGweaver, which is fast enough to be applied to sequences megabases in length. With the aid of a large computer cluster, ARGweaver can be used to sample full ARGs for entire mammalian genome sequences. We show that ARGweaver performs well in simulation experiments and demonstrate that it can be used to provide new insights about both demographic processes and natural selection when applied to real human genome sequence data.