Evaluation of methods for the inference of ancestral recombination graphs

The ancestral recombination graph (ARG) is a structure that describes the joint genealogies of sampled DNA sequences along the genome. Recent computational methods have made impressive progress towards scalably estimating whole-genome genealogies. In addition to inferring the ARG, some of these methods can also provide ARGs sampled from a defined posterior distribution. Obtaining good samples of ARGs is crucial for quantifying statistical uncertainty and for estimating population genetic parameters such as effective population size, mutation rate, and allele age. Here, we use simulations to benchmark three popular ARG inference programs: ARGweaver, Relate, and tsdate. We use neutral coalescent simulations to 1) compare the true coalescence times to the inferred times at each locus; 2) compare the distribution of coalescence times across all loci to the expected exponential distribution; 3) evaluate whether the sampled coalescence times have the properties expected of a valid posterior distribution. We find that inferred coalescence times at each locus are more accurate in ARGweaver and Relate than in tsdate. However, all three methods tend to overestimate small coalescence times and underestimate large ones. Lastly, the posterior distribution of ARGweaver is closer to the expected posterior distribution than Relate's, but this higher accuracy comes at a substantial trade-off in scalability. The best choice of method will depend on the number and length of input sequences and on the goal of downstream analyses, and we provide guidelines for the best practices.

Recoverability of Ancestral Recombination Graph Topologies

Recombination is a powerful evolutionary process that shapes the genetic diversity observed in the populations of many species. Reconstructing genealogies in the presence of recombination from sequencing data is a very challenging problem, as this relies on mutations having occurred on the correct lineages in order to detect the recombination and resolve the placement of edges in the local trees. We investigate the probability of recovering the true topology of ancestral recombination graphs (ARGs) under the coalescent with recombination and gene conversion. We explore how sample size and mutation rate affect the inherent uncertainty in reconstructed ARGs; this sheds light on the theoretical limitations of ARG reconstruction methods. We illustrate our results using estimates of evolutionary rates for several biological organisms; in particular, we find that for parameter values that are realistic for SARS-CoV-2, the probability of reconstructing genealogies that are close to the truth is low.

An ancestral recombination graph of human, Neanderthal, and Denisovan genomes

Many humans carry genes from Neanderthals, a legacy of past admixture. Existing methods detect this archaic hominin ancestry within human genomes using patterns of linkage disequilibrium or direct comparison to Neanderthal genomes. Each of these methods is limited in sensitivity and scalability. We describe a new ancestral recombination graph inference algorithm that scales to large genome-wide datasets and demonstrate its accuracy on real and simulated data. We then generate a genome-wide ancestral recombination graph including human and archaic hominin genomes. From this, we generate a map within human genomes of archaic ancestry and of genomic regions not shared with archaic hominins either by admixture or incomplete lineage sorting. We find that only 1.5 to 7% of the modern human genome is uniquely human. We also find evidence of multiple bursts of adaptive changes specific to modern humans within the past 600,000 years involving genes related to brain development and function.

SIA: Selection Inference Using the Ancestral Recombination Graph

Detecting signals of selection from genomic data is a central problem in population genetics. Coupling the rich information in the ancestral recombination graph (ARG) with a powerful and scalable deep learning framework, we developed a novel method to detect and quantify positive selection: Selection Inference using the Ancestral recombination graph (SIA). Built on a Long Short-Term Memory (LSTM) architecture, a particular type of a Recurrent Neural Network (RNN), SIA can be trained to explicitly infer a full range of selection coefficients, as well as the allele frequency trajectory and time of selection onset. We benchmarked SIA extensively on simulations under a European human demographic model, and found that it performs as well or better as some of the best available methods, including state-of-the-art machine-learning and ARG-based methods. In addition, we used SIA to estimate selection coefficients at several loci associated with human phenotypes of interest. SIA detected novel signals of selection particular to the European (CEU) population at the MC1R and ABCC11 loci. In addition, it recapitulated signals of selection at the LCT locus and several pigmentation-related genes. Finally, we reanalyzed polymorphism data of a collection of recently radiated southern capuchino seedeater taxa in the genus Sporophila to quantify the strength of selection and improved the power of our previous methods to detect partial soft sweeps. Overall, SIA uses deep learning to leverage the ARG and thereby provides new insight into how selective sweeps shape genomic diversity.

Multiple Loci Selection with Multi-Way Epistasis in Coalescence with Recombination

As studies move into deeper characterization of the impact of selection through non-neutral mutations in whole genome population genetics, modeling for selection becomes crucial. Moreover, epistasis has long been recognized as a significant component in understanding the evolution of complex genetic systems. We present a backward coalescent model, EpiSimRA, that accommodates multiple loci selection, with multi-way (k-way) epistasis for any arbitrary k. Starting from arbitrary extant populations with epistatic sites, we trace the Ancestral Recombination Graph (ARG), sampling relevant recombination and coalescent events. Our framework allows for studying different complex evolutionary scenarios in the presence of selective sweeps, positive and negative selection with multiway epistasis. We also present a forward counterpart of the coalescent model based on a Wright-Fisher (WF) process, which we use as a validation framework, comparing the hallmarks of the ARG between the two. We provide the first framework that allows a nose-to-nose comparison of multiway epistasis in a coalescent simulator with its forward counterpart with respect to the hallmarks of the ARG. We demonstrate, through extensive experiments, that EpiSimRA is consistently superior in terms of performance (seconds vs. hours) in comparison to the forward model without compromising on its accuracy.

Multiple Loci Selection with Multi-way Epistasis in Coalescence with Recombination

As studies move into deeper characterization of the impact of selection through non-neutral mutations in whole genome population genetics, modeling for selection becomes crucial. Moreover, epistasis has long been recognized as a significant component in understanding evolution of complex genetic systems. We present a backward coalescent model EpiSimRA, that builds multiple loci selection, with multiway (k-way) epistasis for any arbitrary k. Starting from arbitrary extant populations with epistatic sites, we trace the Ancestral Recombination Graph (ARG), sampling relevant recombination and coalescent events. Our framework allows for studying different complex evolutionary scenarios in the presence of selective sweeps, positive and negative selection with multiway epistasis. We also present a forward counterpart of the coalescent model based on a Wright-Fisher (WF) process which we use as a validation framework, comparing the hallmarks of the ARG between the two. We provide the first framework that allows a nose-to-nose comparison of multiway epistasis in a coalescent simulator with its forward counterpart with respect to the hallmarks of the ARG. We demonstrate through extensive experiments, that EpiSimRA is consistently superior in term of performance (seconds vs. hours) in comparison to the forward model without compromising on its accuracy.

The distribution of waiting distances in ancestral recombination graphs and its applications

The ancestral recombination graph (ARG) contains the full genealogical information of the sample, and many population genetic inference problems can be solved using inferred or sampled ARGs. In particular, the waiting distance between tree changes along the genome can be used to make inference about the distribution and evolution of recombination rates. To this end, we here derive an analytic expression for the distribution of waiting distances between tree changes under the sequentially Markovian coalescent model and obtain an accurate approximation to the distribution of waiting distances for topology changes. We use these results to show that some of the recently proposed methods for inferring sequences of trees along the genome provide strongly biased distributions of waiting distances. In addition, we provide a correction to an undercounting problem facing all available ARG inference methods, thereby facilitating the use of ARG inference methods to estimate temporal changes in the recombination rate.

Genomic islands of differentiation in a rapid avian radiation have been driven by recent selective sweeps

Numerous studies of emerging species have identified genomic “islands” of elevated differentiation against a background of relative homogeneity. The causes of these islands remain unclear, however, with some signs pointing toward “speciation genes” that locally restrict gene flow and others suggesting selective sweeps that have occurred within nascent species after speciation. Here, we examine this question through the lens of genome sequence data for five species of southern capuchino seedeaters, finch-like birds from South America that have undergone a species radiation during the last ∼50,000 generations. By applying newly developed statistical methods for ancestral recombination graph inference and machine-learning methods for the prediction of selective sweeps, we show that previously identified islands of differentiation in these birds appear to be generally associated with relatively recent, species-specific selective sweeps, most of which are predicted to be soft sweeps acting on standing genetic variation. Many of these sweeps coincide with genes associated with melanin-based variation in plumage, suggesting a prominent role for sexual selection. At the same time, a few loci also exhibit indications of possible selection against gene flow. These observations shed light on the complex manner in which natural selection shapes genome sequences during speciation.

Estimating the time since admixture from phased and unphased molecular data

After admixture, recombination breaks down genomic blocks of contiguous ancestry. The break down of these blocks forms a new ‘molecular’ clock, that ticks at a much faster rate than the mutation clock, enabling accurate dating of admixture events in the recent past. However, existing theory on the break down of these blocks, or the accumulation of delineations between blocks, so called ‘junctions’, has been limited to using regularly spaced markers on phased data. Here, we present an extension to the theory of junctions using the Ancestral Recombination Graph that describes the expected number of junctions for any distribution of markers along the genome. Furthermore, we provide a new framework to infer the time since admixture using unphased data. We demonstrate both the phased and unphased methods on simulated data and show that our new extensions perform much better than previous methods, especially for more ancient admixture times. Lastly, we demonstrate the applicability of our method on an empirical dataset of labcrosses of yeast (Saccharomyces cerevisae) and on two case studies of hybridization in swordtail fish and Populus trees.