Deep convolutional and conditional neural networks for large-scale genomic data generation.

Applications of generative models for genomic data have gained significant momentum in the past few years, with scopes ranging from data characterization to generation of genomic segments and functional sequences. In our previous study, we demonstrated that generative adversarial networks (GANs) and restricted Boltzmann machines (RBMs) can be used to create novel high-quality artificial genomes (AGs) which can preserve the complex characteristics of real genomes such as population structure, linkage disequilibrium and selection signals. However, a major drawback of these models is scalability, since the large feature space of genome-wide data increases computational complexity vastly. To address this issue, we implemented a novel convolutional Wasserstein GAN (WGAN) model along with a novel conditional RBM (CRBM) framework for generating AGs with high SNP number. These networks implicitly learn the varying landscape of haplotypic structure in order to capture complex correlation patterns along the genome and generate a wide diversity of plausible haplotypes. We performed comparative analyses to assess both the quality of these generated haplotypes and the amount of possible privacy leakage from the training data. As the importance of genetic privacy becomes more prevalent, the need for effective privacy protection measures for genomic data increases. We used generative neural networks to create large artificial genome segments which possess many characteristics of real genomes without substantial privacy leakage from the training dataset. In the near future, with further improvements in haplotype quality and privacy preservation, large-scale artificial genome databases can be assembled to provide easily accessible surrogates of real databases, allowing researchers to conduct studies with diverse genomic data within a safe ethical framework in terms of donor privacy.

Simulation-based Benchmarking of Ancient Haplotype Inference for Detecting Population Structure.

Paleogenomic data has informed us about the movements, growth, and relationships of ancient populations. It has also given us context for medically relevant adaptations that appear in present-day humans due to introgression from other hominids, and it continues to help us characterize the evolutionary history of humans. However, ancient DNA (aDNA) presents several practical challenges as various factors such as deamination, high fragmentation, environmental contamination of aDNA, and low amounts of recoverable endogenous DNA, make aDNA recovery and analysis more difficult than modern DNA. Most studies with aDNA leverage only SNP data, and only a few studies have made inferences on human demographic history based on haplotype data, possibly because haplotype estimation (or phasing) has not yet been systematically evaluated in the context of aDNA. Here, we evaluate how the unique challenges of aDNA can impact phasing quality. We also develop a software tool that simulates aDNA taking into account the features of aDNA as well as the evolutionary history of the population. We measured phasing error as a function of aDNA quality and demographic history, and found that low phasing error is achievable even for very ancient individuals (~ 400 generations in the past) as long as contamination and read depth are adequate. Our results show that population splits or bottleneck events occurring between the reference and phased populations affect phasing quality, with bottlenecks resulting in the highest average error rates. Finally, we found that using estimated haplotypes, even if not completely accurate, is superior to using the simulated genotype data when reconstructing changes in population structure after population splits between present-day and ancient populations.

The MUC19 gene in Denisovans, Neanderthals, and Modern Humans: An Evolutionary History of Recurrent Introgression and Natural Selection.

All humans carry a small fraction of archaic ancestry across the genome, the legacy of gene flow from Neanderthals, Denisovans, and other hominids into the ancestors of modern humans. While the effects of Neanderthal ancestry on human fitness and health have been explored more thoroughly, there are fewer examples of adaptive introgression of Denisovan variants. Here, we study the gene MUC19, for which some modern humans carry a Denisovan-like haplotype. MUC19 is a mucin, a glycoprotein that forms gels with various biological functions, from lubrication to immunity. We find the diagnostic variants for the Denisovan-like MUC19 haplotype at high frequencies in admixed Latin American individuals among global population, and at highest frequency in 23 ancient Indigenous American individuals, all predating population admixture with Europeans and Africans. We find that some Neanderthals--Vindija and Chagyrskaya--carry the Denisovan-like MUC19 haplotype, and that it was likely introgressed into human populations through Neanderthal introgression rather than Denisovan introgression. Finally, we find that the Denisovan-like MUC19 haplotype carries a higher copy number of a 30 base-pair variable number tandem repeat relative to the Human-like haplotype, and that copy numbers of this repeat are exceedingly high in American populations. Our results suggest that the Denisovan-like MUC19 haplotype served as the raw genetic material for positive selection as American populations adapted to novel environments during their movement from Beringia into North and then South America.

A Quantitative Theory for Genomic Offset Statistics.

Genomic offset statistics predict the maladaptation of populations to rapid habitat alteration based on association of genotypes with environmental variation. Despite substantial evidence for empirical validity, genomic offset statistics have well-identified limitations, and lack a theory that would facilitate interpretations of predicted values. Here, we clarified the theoretical relationships between genomic offset statistics and unobserved fitness traits controlled by environmentally selected loci and proposed a geometric measure to predict fitness after rapid change in local environment. The predictions of our theory were verified in computer simulations and in empirical data on African pearl millet (Cenchrus americanus) obtained from a common garden experiment. Our results proposed a unified perspective on genomic offset statistics and provided a theoretical foundation necessary when considering their potential application in conservation management in the face of environmental change.

An Overview of Deep Generative Models in Functional and Evolutionary Genomics.

Following the widespread use of deep learning for genomics, deep generative modeling is also becoming a viable methodology for the broad field. Deep generative models (DGMs) can learn the complex structure of genomic data and allow researchers to generate novel genomic instances that retain the real characteristics of the original dataset. Aside from data generation, DGMs can also be used for dimensionality reduction by mapping the data space to a latent space, as well as for prediction tasks via exploitation of this learned mapping or supervised/semi-supervised DGM designs. In this review, we briefly introduce generative modeling and two currently prevailing architectures, we present conceptual applications along with notable examples in functional and evolutionary genomics, and we provide our perspective on potential challenges and future directions.

Demographic history and genetic structure in pre-Hispanic Central Mexico.

Aridoamerica and Mesoamerica are two distinct cultural areas in northern and central Mexico, respectively, that hosted numerous pre-Hispanic civilizations between 2500 BCE and 1521 CE. The division between these regions shifted southward because of severe droughts ~1100 years ago, which allegedly drove a population replacement in central Mexico by Aridoamerican peoples. In this study, we present shotgun genome-wide data from 12 individuals and 27 mitochondrial genomes from eight pre-Hispanic archaeological sites across Mexico, including two at the shifting border of Aridoamerica and Mesoamerica. We find population continuity that spans the climate change episode and a broad preservation of the genetic structure across present-day Mexico for the past 2300 years. Lastly, we identify a contribution to pre-Hispanic populations of northern and central Mexico from two ancient unsampled "ghost" populations.

Cultural transmission of reproductive success impacts genomic diversity, coalescent tree topologies, and demographic inferences.

Cultural transmission of reproductive success has been observed in many human populations as well as other animals. Cultural transmission of reproductive success consists of a positive correlation of nongenetic origin between the progeny size of parents and children. This correlation can result from various factors, such as the social influence of parents on their children, the increase of children's survival through allocare from uncles and aunts, or the transmission of resources. Here, we study the evolution of genomic diversity over time under cultural transmission of reproductive success. Cultural transmission of reproductive success has a threefold impact on population genetics: (1) the effective population size decreases when cultural transmission of reproductive success starts, mimicking a population contraction, and increases back to its original value when cultural transmission of reproductive success stops; (2) coalescent tree topologies are distorted under cultural transmission of reproductive success, with higher imbalance and a higher number of polytomies; and (3) branch lengths are reduced nonhomogenously, with a higher impact on older branches. Under long-lasting cultural transmission of reproductive success, the effective population size stabilizes but the distortion of tree topology and the nonhomogenous branch length reduction remain, yielding U-shaped site frequency spectra under a constant population size. We show that this yields a bias in site frequency spectra-based demographic inference. Considering that cultural transmission of reproductive success was detected in numerous human and animal populations worldwide, one should be cautious because inferring population past histories from genomic data can be biased by this cultural process.

dnadna: a deep learning framework for population genetics inference.

MOTIVATION: We present dnadna, a flexible python-based software for deep learning inference in population genetics. It is task-agnostic and aims at facilitating the development, reproducibility, dissemination and re-usability of neural networks designed for population genetic data. RESULTS: dnadna defines multiple user-friendly workflows. First, users can implement new architectures and tasks, while benefiting from dnadna utility functions, training procedure and test environment, which saves time and decreases the likelihood of bugs. Second, the implemented networks can be re-optimized based on user-specified training sets and/or tasks. Newly implemented architectures and pre-trained networks are easily shareable with the community for further benchmarking or other applications. Finally, users can apply pre-trained networks in order to predict evolutionary history from alternative real or simulated genetic datasets, without requiring extensive knowledge in deep learning or coding in general. dnadna comes with a peer-reviewed, exchangeable neural network, allowing demographic inference from SNP data, that can be used directly or retrained to solve other tasks. Toy networks are also available to ease the exploration of the software, and we expect that the range of available architectures will keep expanding thanks to community contributions. AVAILABILITY AND IMPLEMENTATION: dnadna is a Python (≥3.7) package, its repository is available at gitlab.com/mlgenetics/dnadna and its associated documentation at mlgenetics.gitlab.io/dnadna/.

Creating artificial human genomes using generative neural networks.

Generative models have shown breakthroughs in a wide spectrum of domains due to recent advancements in machine learning algorithms and increased computational power. Despite these impressive achievements, the ability of generative models to create realistic synthetic data is still under-exploited in genetics and absent from population genetics. Yet a known limitation in the field is the reduced access to many genetic databases due to concerns about violations of individual privacy, although they would provide a rich resource for data mining and integration towards advancing genetic studies. In this study, we demonstrated that deep generative adversarial networks (GANs) and restricted Boltzmann machines (RBMs) can be trained to learn the complex distributions of real genomic datasets and generate novel high-quality artificial genomes (AGs) with none to little privacy loss. We show that our generated AGs replicate characteristics of the source dataset such as allele frequencies, linkage disequilibrium, pairwise haplotype distances and population structure. Moreover, they can also inherit complex features such as signals of selection. To illustrate the promising outcomes of our method, we showed that imputation quality for low frequency alleles can be improved by data augmentation to reference panels with AGs and that the RBM latent space provides a relevant encoding of the data, hence allowing further exploration of the reference dataset and features for solving supervised tasks. Generative models and AGs have the potential to become valuable assets in genetic studies by providing a rich yet compact representation of existing genomes and high-quality, easy-access and anonymous alternatives for private databases.

Deep learning for population size history inference: Design, comparison and combination with approximate Bayesian computation.

For the past decades, simulation-based likelihood-free inference methods have enabled researchers to address numerous population genetics problems. As the richness and amount of simulated and real genetic data keep increasing, the field has a strong opportunity to tackle tasks that current methods hardly solve. However, high data dimensionality forces most methods to summarize large genomic data sets into a relatively small number of handcrafted features (summary statistics). Here, we propose an alternative to summary statistics, based on the automatic extraction of relevant information using deep learning techniques. Specifically, we design artificial neural networks (ANNs) that take as input single nucleotide polymorphic sites (SNPs) found in individuals sampled from a single population and infer the past effective population size history. First, we provide guidelines to construct artificial neural networks that comply with the intrinsic properties of SNP data such as invariance to permutation of haplotypes, long scale interactions between SNPs and variable genomic length. Thanks to a Bayesian hyperparameter optimization procedure, we evaluate the performance of multiple networks and compare them to well-established methods like Approximate Bayesian Computation (ABC). Even without the expert knowledge of summary statistics, our approach compares fairly well to an ABC approach based on handcrafted features. Furthermore, we show that combining deep learning and ABC can improve performance while taking advantage of both frameworks. Finally, we apply our approach to reconstruct the effective population size history of cattle breed populations.

Factor analysis of ancient population genomic samples.

The recent years have seen a growing number of studies investigating evolutionary questions using ancient DNA. To address these questions, one of the most frequently-used method is principal component analysis (PCA). When PCA is applied to temporal samples, the sample dates are, however, ignored during analysis, leading to imperfect representations of samples in PC plots. Here, we present a factor analysis (FA) method in which individual scores are corrected for the effect of allele frequency drift over time. We obtained exact solutions for the estimates of corrected factors, and we provided a fast algorithm for their computation. Using computer simulations and ancient European samples, we compared geometric representations obtained from FA with PCA and with ancestry estimation programs. In admixture analyses, FA estimates agreed with tree-based statistics, and they were more accurate than those obtained from PCA projections and from ancestry estimation programs. A great advantage of FA over existing approaches is to improve descriptive analyses of ancient DNA samples without requiring inclusion of outgroup or present-day samples.

Differences in local population history at the finest level: the case of the Estonian population.

Several recent studies detected fine-scale genetic structure in human populations. Hence, groups conventionally treated as single populations harbour significant variation in terms of allele frequencies and patterns of haplotype sharing. It has been shown that these findings should be considered when performing studies of genetic associations and natural selection, especially when dealing with polygenic phenotypes. However, there is little understanding of the practical effects of such genetic structure on demography reconstructions and selection scans when focusing on recent population history. Here we tested the impact of population structure on such inferences using high-coverage (~30×) genome sequences of 2305 Estonians. We show that different regions of Estonia differ in both effective population size dynamics and signatures of natural selection. By analyzing identity-by-descent segments we also reveal that some Estonian regions exhibit evidence of a bottleneck 10-15 generations ago reflecting sequential episodes of wars, plague and famine, although this signal is virtually undetected when treating Estonia as a single population. Besides that, we provide a framework for relating effective population size estimated from genetic data to actual census size and validate it on the Estonian population. This approach may be widely used both to cross-check estimates based on historical sources as well as to get insight into times and/or regions with no other information available. Our results suggest that the history of human populations within the last few millennia can be highly region specific and cannot be properly studied without taking local genetic structure into account.

An ABC Method for Whole-Genome Sequence Data: Inferring Paleolithic and Neolithic Human Expansions.

Species generally undergo a complex demographic history consisting, in particular, of multiple changes in population size. Genome-wide sequencing data are potentially highly informative for reconstructing this demographic history. A crucial point is to extract the relevant information from these very large data sets. Here, we design an approach for inferring past demographic events from a moderate number of fully sequenced genomes. Our new approach uses Approximate Bayesian Computation, a simulation-based statistical framework that allows 1) identifying the best demographic scenario among several competing scenarios and 2) estimating the best-fitting parameters under the chosen scenario. Approximate Bayesian Computation relies on the computation of summary statistics. Using a cross-validation approach, we show that statistics such as the lengths of haplotypes shared between individuals, or the decay of linkage disequilibrium with distance, can be combined with classical statistics (e.g., heterozygosity and Tajima's D) to accurately infer complex demographic scenarios including bottlenecks and expansion periods. We also demonstrate the importance of simultaneously estimating the genotyping error rate. Applying our method on genome-wide human-sequence databases, we finally show that a model consisting in a bottleneck followed by a Paleolithic and a Neolithic expansion is the most relevant for Eurasian populations.

Determination of genetic relatedness from low-coverage human genome sequences using pedigree simulations.

We develop and evaluate methods for inferring relatedness among individuals from low-coverage DNA sequences of their genomes, with particular emphasis on sequences obtained from fossil remains. We suggest the major factors complicating the determination of relatedness among ancient individuals are sequencing depth, the number of overlapping sites, the sequencing error rate and the presence of contamination from present-day genetic sources. We develop a theoretical model that facilitates the exploration of these factors and their relative effects, via measurement of pairwise genetic distances, without calling genotypes, and determine the power to infer relatedness under various scenarios of varying sequencing depth, present-day contamination and sequencing error. The model is validated by a simulation study as well as the analysis of aligned sequences from present-day human genomes. We then apply the method to the recently published genome sequences of ancient Europeans, developing a statistical treatment to determine confidence in assigned relatedness that is, in some cases, more precise than previously reported. As the majority of ancient specimens are from animals, this method would be applicable to investigate kinship in nonhuman remains. The developed software grups (Genetic Relatedness Using Pedigree Simulations) is implemented in Python and freely available.

Inferring Population Size History from Large Samples of Genome-Wide Molecular Data - An Approximate Bayesian Computation Approach.

Inferring the ancestral dynamics of effective population size is a long-standing question in population genetics, which can now be tackled much more accurately thanks to the massive genomic data available in many species. Several promising methods that take advantage of whole-genome sequences have been recently developed in this context. However, they can only be applied to rather small samples, which limits their ability to estimate recent population size history. Besides, they can be very sensitive to sequencing or phasing errors. Here we introduce a new approximate Bayesian computation approach named PopSizeABC that allows estimating the evolution of the effective population size through time, using a large sample of complete genomes. This sample is summarized using the folded allele frequency spectrum and the average zygotic linkage disequilibrium at different bins of physical distance, two classes of statistics that are widely used in population genetics and can be easily computed from unphased and unpolarized SNP data. Our approach provides accurate estimations of past population sizes, from the very first generations before present back to the expected time to the most recent common ancestor of the sample, as shown by simulations under a wide range of demographic scenarios. When applied to samples of 15 or 25 complete genomes in four cattle breeds (Angus, Fleckvieh, Holstein and Jersey), PopSizeABC revealed a series of population declines, related to historical events such as domestication or modern breed creation. We further highlight that our approach is robust to sequencing errors, provided summary statistics are computed from SNPs with common alleles.

Genome sequence of a 45,000-year-old modern human from western Siberia.

We present the high-quality genome sequence of a ∼45,000-year-old modern human male from Siberia. This individual derives from a population that lived before-or simultaneously with-the separation of the populations in western and eastern Eurasia and carries a similar amount of Neanderthal ancestry as present-day Eurasians. However, the genomic segments of Neanderthal ancestry are substantially longer than those observed in present-day individuals, indicating that Neanderthal gene flow into the ancestors of this individual occurred 7,000-13,000 years before he lived. We estimate an autosomal mutation rate of 0.4 × 10(-9) to 0.6 × 10(-9) per site per year, a Y chromosomal mutation rate of 0.7 × 10(-9) to 0.9 × 10(-9) per site per year based on the additional substitutions that have occurred in present-day non-Africans compared to this genome, and a mitochondrial mutation rate of 1.8 × 10(-8) to 3.2 × 10(-8) per site per year based on the age of the bone.

Insights into persistence mechanisms of a zoonotic virus in bat colonies using a multispecies metapopulation model.

Rabies is a worldwide zoonosis resulting from Lyssavirus infection. In Europe, Eptesicus serotinus is the most frequently reported bat species infected with Lyssavirus, and thus considered to be the reservoir of European bat Lyssavirus type 1 (EBLV-1). To date, the role of other bat species in EBLV-1 epidemiology and persistence remains unknown. Here, we built an EBLV-1-transmission model based on local observations of a three-cave and four-bat species (Myotis capaccinii, Myotis myotis, Miniopterus schreibersii, Rhinolophus ferrumequinum) system in the Balearic Islands, for which a 1995-2011 serological dataset indicated the continuous presence of EBLV-1. Eptesicus serotinus was never observed in the system during the 16-year follow-up and therefore was not included in the model. We used the model to explore virus persistence mechanisms and to assess the importance of each bat species in the transmission dynamics. We found that EBLV-1 could not be sustained if transmission between M. schreibersii and other bat species was eliminated, suggesting that this species serves as a regional reservoir. Global sensitivity analysis using Sobol's method revealed that following the rate of autumn-winter infectious contacts, M. schreibersii's incubation- and immune-period durations, but not the infectious period length, were the most relevant factors driving virus persistence.

The complete genome sequence of a Neanderthal from the Altai Mountains.

We present a high-quality genome sequence of a Neanderthal woman from Siberia. We show that her parents were related at the level of half-siblings and that mating among close relatives was common among her recent ancestors. We also sequenced the genome of a Neanderthal from the Caucasus to low coverage. An analysis of the relationships and population history of available archaic genomes and 25 present-day human genomes shows that several gene flow events occurred among Neanderthals, Denisovans and early modern humans, possibly including gene flow into Denisovans from an unknown archaic group. Thus, interbreeding, albeit of low magnitude, occurred among many hominin groups in the Late Pleistocene. In addition, the high-quality Neanderthal genome allows us to establish a definitive list of substitutions that became fixed in modern humans after their separation from the ancestors of Neanderthals and Denisovans.

Genomic evidence for island population conversion resolves conflicting theories of polar bear evolution.

Despite extensive genetic analysis, the evolutionary relationship between polar bears (Ursus maritimus) and brown bears (U. arctos) remains unclear. The two most recent comprehensive reports indicate a recent divergence with little subsequent admixture or a much more ancient divergence followed by extensive admixture. At the center of this controversy are the Alaskan ABC Islands brown bears that show evidence of shared ancestry with polar bears. We present an analysis of genome-wide sequence data for seven polar bears, one ABC Islands brown bear, one mainland Alaskan brown bear, and a black bear (U. americanus), plus recently published datasets from other bears. Surprisingly, we find clear evidence for gene flow from polar bears into ABC Islands brown bears but no evidence of gene flow from brown bears into polar bears. Importantly, while polar bears contributed <1% of the autosomal genome of the ABC Islands brown bear, they contributed 6.5% of the X chromosome. The magnitude of sex-biased polar bear ancestry and the clear direction of gene flow suggest a model wherein the enigmatic ABC Island brown bears are the descendants of a polar bear population that was gradually converted into brown bears via male-dominated brown bear admixture. We present a model that reconciles heretofore conflicting genetic observations. We posit that the enigmatic ABC Islands brown bears derive from a population of polar bears likely stranded by the receding ice at the end of the last glacial period. Since then, male brown bear migration onto the island has gradually converted these bears into an admixed population whose phenotype and genotype are principally brown bear, except at mtDNA and X-linked loci. This process of genome erosion and conversion may be a common outcome when climate change or other forces cause a population to become isolated and then overrun by species with which it can hybridize.

Higher levels of neanderthal ancestry in East Asians than in Europeans.

Neanderthals were a group of archaic hominins that occupied most of Europe and parts of Western Asia from ∼30,000 to 300,000 years ago (KYA). They coexisted with modern humans during part of this time. Previous genetic analyses that compared a draft sequence of the Neanderthal genome with genomes of several modern humans concluded that Neanderthals made a small (1-4%) contribution to the gene pools of all non-African populations. This observation was consistent with a single episode of admixture from Neanderthals into the ancestors of all non-Africans when the two groups coexisted in the Middle East 50-80 KYA. We examined the relationship between Neanderthals and modern humans in greater detail by applying two complementary methods to the published draft Neanderthal genome and an expanded set of high-coverage modern human genome sequences. We find that, consistent with the recent finding of Meyer et al. (2012), Neanderthals contributed more DNA to modern East Asians than to modern Europeans. Furthermore we find that the Maasai of East Africa have a small but significant fraction of Neanderthal DNA. Because our analysis is of several genomic samples from each modern human population considered, we are able to document the extent of variation in Neanderthal ancestry within and among populations. Our results combined with those previously published show that a more complex model of admixture between Neanderthals and modern humans is necessary to account for the different levels of Neanderthal ancestry among human populations. In particular, at least some Neanderthal-modern human admixture must postdate the separation of the ancestors of modern European and modern East Asian populations.