Recommendations for improving statistical inference in population genomics.

The field of population genomics has grown rapidly in response to the recent advent of affordable, large-scale sequencing technologies. As opposed to the situation during the majority of the 20th century, in which the development of theoretical and statistical population genetic insights outpaced the generation of data to which they could be applied, genomic data are now being produced at a far greater rate than they can be meaningfully analyzed and interpreted. With this wealth of data has come a tendency to focus on fitting specific (and often rather idiosyncratic) models to data, at the expense of a careful exploration of the range of possible underlying evolutionary processes. For example, the approach of directly investigating models of adaptive evolution in each newly sequenced population or species often neglects the fact that a thorough characterization of ubiquitous nonadaptive processes is a prerequisite for accurate inference. We here describe the perils of these tendencies, present our consensus views on current best practices in population genomic data analysis, and highlight areas of statistical inference and theory that are in need of further attention. Thereby, we argue for the importance of defining a biologically relevant baseline model tuned to the details of each new analysis, of skepticism and scrutiny in interpreting model fitting results, and of carefully defining addressable hypotheses and underlying uncertainties.

Soft selective sweeps: Addressing new definitions, evaluating competing models, and interpreting empirical outliers.

The ability to accurately identify and quantify genetic signatures associated with soft selective sweeps based on patterns of nucleotide variation has remained controversial. We here provide counter viewpoints to recent publications in PLOS Genetics that have argued not only for the statistical identifiability of soft selective sweeps, but also for their pervasive evolutionary role in both Drosophila and HIV populations. We present evidence that these claims owe to a lack of consideration of competing evolutionary models, unjustified interpretations of empirical outliers, as well as to new definitions of the processes themselves. Our results highlight the dangers of fitting evolutionary models based on hypothesized and episodic processes without properly first considering common processes and, more generally, of the tendency in certain research areas to view pervasive positive selection as a foregone conclusion.

Sweepstakes reproduction facilitates rapid adaptation in highly fecund populations.

Adaptation enables natural populations to survive in a changing environment. Understanding the mechanics of adaptation is therefore crucial for learning about the evolution and ecology of natural populations. We focus on the impact of random sweepstakes on selection in highly fecund haploid and diploid populations partitioned into two genetic types, with one type conferring selective advantage. For the diploid populations, we incorporate various dominance mechanisms. We assume that the populations may experience recurrent bottlenecks. In random sweepstakes, the distribution of individual recruitment success is highly skewed, resulting in a huge variance in the number of offspring contributed by the individuals present in any given generation. Using computer simulations, we investigate the joint effects of random sweepstakes, recurrent bottlenecks and dominance mechanisms on selection. In our framework, bottlenecks allow random sweepstakes to have an effect on the time to fixation, and in diploid populations, the effect of random sweepstakes depends on the dominance mechanism. We describe selective sweepstakes that are approximated by recurrent sweeps of strongly beneficial allelic types arising by mutation. We demonstrate that both types of sweepstakes reproduction may facilitate rapid adaptation (as defined based on the average time to fixation of a type conferring selective advantage conditioned on fixation of the type). However, whether random sweepstakes cause rapid adaptation depends also on their interactions with bottlenecks and dominance mechanisms. Finally, we review a case study in which a model of recurrent sweeps is shown to essentially explain population genomic data from Atlantic cod.

Polygenic Adaptation in a Population of Finite Size.

Polygenic adaptation in response to selection on quantitative traits has become an important topic in evolutionary biology. Here we review the recent literature on models of polygenic adaptation. In particular, we focus on a model that includes mutation and both directional and stabilizing selection on a highly polygenic trait in a population of finite size (thus experiencing random genetic drift). Assuming that a sudden environmental shift of the fitness optimum occurs while the population is in a stochastic equilibrium, we analyze the adaptation of the trait to the new optimum. When the shift is not too large relative to the equilibrium genetic variance and this variance is determined by loci with mostly small effects, the approach of the mean phenotype to the optimum can be approximated by a rapid exponential process (whose rate is proportional to the genetic variance). During this rapid phase the underlying changes to allele frequencies, however, may depend strongly on genetic drift. While trait-increasing alleles with intermediate equilibrium frequencies are dominated by selection and contribute positively to changes of the trait mean (i.e., are aligned with the direction of the optimum shift), alleles with low or high equilibrium frequencies show more of a random dynamics, which is expected when drift is dominating. A strong effect of drift is also predicted for population size bottlenecks. Our simulations show that the presence of a bottleneck results in a larger deviation of the population mean of the trait from the fitness optimum, which suggests that more loci experience the influence of drift.

Neutral genomic signatures of host-parasite coevolution.

BACKGROUND: Coevolution is a selective process of reciprocal adaptation in hosts and parasites or in mutualistic symbionts. Classic population genetics theory predicts the signatures of selection at the interacting loci of both species, but not the neutral genome-wide polymorphism patterns. To bridge this gap, we build an eco-evolutionary model, where neutral genomic changes over time are driven by a single selected locus in hosts and parasites via a simple biallelic gene-for-gene or matching-allele interaction. This coevolutionary process may lead to cyclic changes in the sizes of the interacting populations. RESULTS: We investigate if and when these changes can be observed in the site frequency spectrum of neutral polymorphisms from host and parasite full genome data. We show that changes of the host population size are too smooth to be observable in its polymorphism pattern over the course of time. Conversely, the parasite population may undergo a series of strong bottlenecks occurring on a slower relative time scale, which may lead to observable changes in a time series sample. We also extend our results to cases with 1) several parasites per host accelerating relative time, and 2) multiple parasite generations per host generation slowing down rescaled time. CONCLUSIONS: Our results show that time series sampling of host and parasite populations with full genome data are crucial to understand if and how coevolution occurs. This model provides therefore a framework to interpret and draw inference from genome-wide polymorphism data of interacting species.

Modes of Rapid Polygenic Adaptation.

Many experimental and field studies have shown that adaptation can occur very rapidly. Two qualitatively different modes of fast adaptation have been proposed: selective sweeps wherein large shifts in the allele frequencies occur at a few loci and evolution via small changes in the allele frequencies at many loci. Although the first process has been thoroughly investigated within the framework of population genetics, the latter is based on quantitative genetics and is much less understood. Here we summarize results from our recent theoretical studies of a quantitative genetic model of polygenic adaptation that makes explicit reference to population genetics to bridge the gap between the two frameworks. Our key results are that polygenic adaptation may be a rapid process and can proceed via subtle or dramatic changes in the allele frequency depending on the sizes of the phenotypic effects relative to a threshold value. We also discuss how the signals of polygenic selection may be detected in the genome. Although powerful methods are available to identify signatures of selective sweeps at loci controlling quantitative traits, the development of statistical tests for detecting small shifts of allele frequencies at quantitative trait loci is still in its infancy.

Evolution of highly fecund haploid populations.

We consider a model of viability selection in a highly fecund haploid population with sweepstakes reproduction. We use simulations to estimate the time until the allelic type with highest fitness has reached high frequency in a finite population. We compare the time between two reproduction modes of high and low fecundity. We also consider the probability that the allelic type with highest fitness is lost from the population before reaching high frequency. Our simulation results indicate that highly fecund populations can evolve faster (in some cases much faster) than populations of low fecundity. However, high fecundity and sweepstakes reproduction also confer much higher risk of losing the allelic type with highest fitness from the population by chance. The impact of selection on driving alleles to high frequency varies depending on the trait value conferring highest fitness; in some cases the effect of selection can hardly be detected.

A genome-wide scan for genes under balancing selection in Drosophila melanogaster.

BACKGROUND: In the history of population genetics balancing selection has been considered as an important evolutionary force, yet until today little is known about its abundance and its effect on patterns of genetic diversity. Several well-known examples of balancing selection have been reported from humans, mice, plants, and parasites. However, only very few systematic studies have been carried out to detect genes under balancing selection. We performed a genome scan in Drosophila melanogaster to find signatures of balancing selection in a derived (European) and an ancestral (African) population. We screened a total of 34 genomes searching for regions of high genetic diversity and an excess of SNPs with intermediate frequency. RESULTS: In total, we found 183 candidate genes: 141 in the European population and 45 in the African one, with only three genes shared between both populations. Most differences between both populations were observed on the X chromosome, though this might be partly due to false positives. Functionally, we find an overrepresentation of genes involved in neuronal development and circadian rhythm. Furthermore, some of the top genes we identified are involved in innate immunity. CONCLUSION: Our results revealed evidence of genes under balancing selection in European and African populations. More candidate genes have been found in the European population. They are involved in several different functions.

Canalization of gene expression is a major signature of regulatory cold adaptation in temperate Drosophila melanogaster.

BACKGROUND: Transcriptome analysis may provide means to investigate the underlying genetic causes of shared and divergent phenotypes in different populations and help to identify potential targets of adaptive evolution. Applying RNA sequencing to whole male Drosophila melanogaster from the ancestral tropical African environment and a very recently colonized cold-temperate European environment at both standard laboratory conditions and following a cold shock, we seek to uncover the transcriptional basis of cold adaptation. RESULTS: In both the ancestral and the derived populations, the predominant characteristic of the cold shock response is the swift and massive upregulation of heat shock proteins and other chaperones. Although we find ~25 % of the genome to be differentially expressed following a cold shock, only relatively few genes (n = 16) are up- or down-regulated in a population-specific way. Intriguingly, 14 of these 16 genes show a greater degree of differential expression in the African population. Likewise, there is an excess of genes with particularly strong cold-induced changes in expression in Africa on a genome-wide scale. CONCLUSIONS: The analysis of the transcriptional cold shock response most prominently reveals an upregulation of components of a general stress response, which is conserved over many taxa and triggered by a plethora of stressors. Despite the overall response being fairly similar in both populations, there is a definite excess of genes with a strong cold-induced fold-change in Africa. This is consistent with a detrimental deregulation or an overshooting stress response. Thus, the canalization of European gene expression might be responsible for the increased cold tolerance of European flies.

North-South Colonization Associated with Local Adaptation of the Wild Tomato Species Solanum chilense.

After colonization population sizes may vary across the species range depending on environmental conditions and following colonizations. An interesting question is whether local adaptation occurs more frequently in large ancestral populations or in small derived populations. A higher number of new mutations and a lower effect of genetic drift should favor selection in large populations, whereas small derived populations may require an initial local adaptation event to facilitate the colonization of new habitats. Wild tomatoes are native to a broad range of different habitats characterized by variable abiotic conditions in South America, and represent an ideal system to study this interplay between demography and natural selection. Population genetic analyses and statistical inference of past demography were conducted on pooled-sequencing data from 30 genes (8,080 single nucleotide polymorphisms) from an extensive sampling of 23 Solanum chilense populations over Chile and Peru. We reveal first a north-south colonization associated with relaxed purifying selection in the south as shown by a decrease of genetic variation and an increasing proportion of nonsynonymous polymorphism from north to south, and population substructure with at least four genetic groups. Second, we uncover a dual picture of adaptation consisting of 1) a decreasing proportion of adaptive amino acid substitutions from north to south suggesting that adaptation is favored in large populations, whereas 2) signatures of local adaptation predominantly occur in the smaller populations from the marginal ranges in the south.

Signatures of positive selection: from selective sweeps at individual loci to subtle allele frequency changes in polygenic adaptation.

In the past 15 years, numerous methods have been developed to detect selective sweeps underlying adaptations. These methods are based on relatively simple population genetic models, including one or two loci at which positive directional selection occurs, and one or two marker loci at which the impact of selection on linked neutral variation is quantified. Information about the phenotype under selection is not included in these models (except for fitness). In contrast, in the quantitative genetic models of adaptation, selection acts on one or more phenotypic traits, such that a genotype-phenotype map is required to bridge the gap to population genetics theory. Here I describe the range of population genetic models from selective sweeps in a panmictic population of constant size to evolutionary traffic when simultaneous sweeps at multiple loci interfere, and I also consider the case of polygenic selection characterized by subtle allele frequency shifts at many loci. Furthermore, I present an overview of the statistical tests that have been proposed based on these population genetics models to detect evidence for positive selection in the genome.

Population genetic evidence for cold adaptation in European Drosophila melanogaster populations.

We studied Drosophila melanogaster populations from Europe (the Netherlands and France) and Africa (Rwanda and Zambia) to uncover genetic evidence of adaptation to cold. We present here four lines of evidence for genes involved in cold adaptation from four perspectives: (i) the frequency of SNPs at genes previously known to be associated with chill-coma recovery time (CCRT), startle reflex (SR) and resistance to starvation stress (RSS) vary along environmental gradients and therefore among populations; (ii) SNPs of genes that correlate significantly with latitude and altitude in African and European populations overlap with SNPs that correlate with a latitudinal cline from North America; (iii) at the genomewide level, the top candidate genes are enriched in gene ontology (GO) terms that are related to cold tolerance; (iv) GO enriched terms from North American clinal genes overlap significantly with those from Africa and Europe. Each SNP was tested in 10 independent runs of Bayenv2, using the median Bayes factors to ascertain candidate genes. None of the candidate genes were found close to the breakpoints of cosmopolitan inversions, and only four candidate genes were linked to QTLs related to CCRT. To overcome the limitation that we used only four populations to test correlations with environmental gradients, we performed simulations to estimate the power of our approach for detecting selection. Based on our results, we propose a novel network of genes that is involved in cold adaptation.

Adaptation to low temperatures in the wild tomato species Solanum chilense.

Molecular adaptation to abiotic stresses in plants is a complex process based mainly on the modifications of gene transcriptional activity and the alteration of protein-protein interactions. We used a combination of population genetic, comparative transcriptomic and plant physiology approaches to investigate the mechanisms of adaptation to low temperatures in Solanum chilense populations distributed along Andean altitudinal gradients. We found that plants from all populations have high chilling tolerance, which does not correlate with temperatures in their native habitats. In contrast, tolerance to freezing shows a significant association with altitude and temperature variables. We also observed the differences in expression patterns of cold-response genes between plants from high- and low-altitude populations. These results suggest that genetic adaptations to low temperatures evolved in high-altitude populations of S. chilense. At the transcriptional level, these adaptations may include high levels of constitutive expression of the genes encoding ICE1, the key transcription factor of the cold signalling pathway, and chloroplast ω-3 fatty acid desaturase FAD7. At the sequence level, a signature of selection associated with the adaptation to high altitudes was detected at the C-terminal part of ICE1 encoding the ACT regulatory domain.

A model of compensatory molecular evolution involving multiple sites in RNA molecules.

Consider two sites under compensatory fitness interaction, such as a Watson-Crick base pair in an RNA helix or two interacting residues in a protein. A mutation at any one of these two sites may reduce the fitness of an individual. However, fitness may be restored by the occurrence of a second mutation at the other site. Kimura modeled this process using a two-locus haploid fitness scheme with two alleles at each locus. He predicted that compensatory evolution following this model is very rare unless selection against the deleterious single mutations is weak and linkage between the interacting sites is tight. Here we investigate the question whether the rate of compensatory evolution increases if we take the context of the two directly interacting sites into account. By "context", we mean the effect of neighboring sites in an RNA helix. Interaction between the focal pair of sites under consideration and the context may lead to so-called indirect compensation. Thus, extending Kimura's classical model of compensatory evolution, we study the effects of both direct and indirect compensation on the rate of compensatory evolution. It is shown that the effects of indirect compensation are very strong. We find that recombination does not slow down the rate of compensatory evolution as predicted by the classical model. Instead, compensatory substitutions may be relatively frequent, even if linkage between the focal interacting sites is loose, selection against deleterious mutations is strong, and mutation rate is low. We compare our theoretical results with data on RNA secondary structures from vertebrate introns.

Balancing selection at the tomato RCR3 Guardee gene family maintains variation in strength of pathogen defense.

Coevolution between hosts and pathogens is thought to occur between interacting molecules of both species. This results in the maintenance of genetic diversity at pathogen antigens (or so-called effectors) and host resistance genes such as the major histocompatibility complex (MHC) in mammals or resistance (R) genes in plants. In plant-pathogen interactions, the current paradigm posits that a specific defense response is activated upon recognition of pathogen effectors via interaction with their corresponding R proteins. According to the "Guard-Hypothesis," R proteins (the "guards") can sense modification of target molecules in the host (the "guardees") by pathogen effectors and subsequently trigger the defense response. Multiple studies have reported high genetic diversity at R genes maintained by balancing selection. In contrast, little is known about the evolutionary mechanisms shaping the guardee, which may be subject to contrasting evolutionary forces. Here we show that the evolution of the guardee RCR3 is characterized by gene duplication, frequent gene conversion, and balancing selection in the wild tomato species Solanum peruvianum. Investigating the functional characteristics of 54 natural variants through in vitro and in planta assays, we detected differences in recognition of the pathogen effector through interaction with the guardee, as well as substantial variation in the strength of the defense response. This variation is maintained by balancing selection at each copy of the RCR3 gene. Our analyses pinpoint three amino acid polymorphisms with key functional consequences for the coevolution between the guardee (RCR3) and its guard (Cf-2). We conclude that, in addition to coevolution at the "guardee-effector" interface for pathogen recognition, natural selection acts on the "guard-guardee" interface. Guardee evolution may be governed by a counterbalance between improved activation in the presence and prevention of auto-immune responses in the absence of the corresponding pathogen.

Inference of seed bank parameters in two wild tomato species using ecological and genetic data.

Seed and egg dormancy is a prevalent life-history trait in plants and invertebrates whose storage effect buffers against environmental variability, modulates species extinction in fragmented habitats, and increases genetic variation. Experimental evidence for reliable differences in dormancy over evolutionary scales (e.g., differences in seed banks between sister species) is scarce because complex ecological experiments in the field are needed to measure them. To cope with these difficulties, we developed an approximate Bayesian computation (ABC) framework that integrates ecological information on population census sizes in the priors of the parameters, along with a coalescent model accounting simultaneously for seed banks and spatial genetic structuring of populations. We collected SNP data at seven nuclear loci (over 300 SNPs) using a combination of three spatial sampling schemes: population, pooled, and species-wide samples. We provide evidence for the existence of a seed bank in two wild tomato species (Solanum chilense and Solanum peruvianum) found in western South America. Although accounting for uncertainties in ecological data, we infer for each species (i) the past demography and (ii) ecological parameters, such as the germination rate, migration rates, and minimum number of demes in the metapopulation. The inferred difference in germination rate between the two species may reflect divergent seed dormancy adaptations, in agreement with previous population genetic analyses and the ecology of these two sister species: Seeds spend, on average, a shorter time in the soil in the specialist species (S. chilense) than in the generalist species (S. peruvianum).

A critical assessment of storytelling: gene ontology categories and the importance of validating genomic scans.

In the age of whole-genome population genetics, so-called genomic scan studies often conclude with a long list of putatively selected loci. These lists are then further scrutinized to annotate these regions by gene function, corresponding biological processes, expression levels, or gene networks. Such annotations are often used to assess and/or verify the validity of the genome scan and the statistical methods that have been used to perform the analyses. Furthermore, these results are frequently considered to validate "true-positives" if the identified regions make biological sense a posteriori. Here, we show that this approach can be potentially misleading. By simulating neutral evolutionary histories, we demonstrate that it is possible not only to obtain an extremely high false-positive rate but also to make biological sense out of the false-positives and construct a sensible biological narrative. Results are compared with a recent polymorphism data set from Drosophila melanogaster.

Trans-species polymorphism and allele-specific expression in the CBF gene family of wild tomatoes.

Abiotic stresses such as drought, extreme temperatures, and salinity have a strong impact on plant adaptation. They act as selective forces on plant physiology and morphology. These selective pressures leave characteristic footprints that can be detected at the DNA sequence level using population genetic tools. On the basis of a candidate gene approach, we investigated signatures of adaptation in two wild tomato species, Solanum peruvianum and S. chilense. These species are native to western South America and constitute a model system for studying adaptation, due to their ability to colonize diverse habitats and the available genetic resources. We have determined the selective forces acting on the C-repeat binding factor (CBF) gene family, which consists of three genes, and is known to be involved in tolerance to abiotic stresses, in particular in cold tolerance. We also analyzed the expression pattern of these genes after drought and cold stresses. We found that CBF3 evolves under very strong purifying selection, CBF2 is under balancing selection in some populations of both species (S. peruvianum/Quicacha and S. chilense/Nazca) maintaining a trans-species polymorphism, and CBF1 is a pseudogene. In contrast to previous studies of cultivated tomatoes showing that only CBF1 was cold induced, we found that all three CBF genes are cold induced in wild tomatoes. All three genes are also drought induced. CBF2 exhibits an allele-specific expression pattern associated with the trans-species polymorphism.

Selective constraints in conserved folded RNAs of drosophilid and hominid genomes.

Small noncoding RNAs as well as folded RNA structures in genic regions are crucial for many cellular processes. They are involved in posttranscriptional gene regulation (microRNAs), RNA modification (small nucleolar RNAs), regulation of splicing, correct localization of proteins, and many other processes. In most cases, a distinct secondary structure of the molecule is necessary for its correct function. Hence, selection should act to retain the structure of the molecule, although the underlying sequence is allowed to vary. Here, we present the first genome-wide estimates of selective constraints in folded RNA molecules in the nuclear genomes of drosophilids and hominids. In comparison to putatively neutrally evolving sites, we observe substantially reduced rates of substitutions at paired and unpaired sites of folded molecules. We estimated evolutionary constraints to be in the ranges of (0.974,0.991) and (0.895,1.000) for paired nucleotides in drosophilids and hominids, respectively. These values are significantly higher than for constraints at nonsynonymous sites of protein-coding genes in both genera. Nonetheless, valleys of only moderately reduced fitness (s ≈ 10(-4)) are sufficient to generate the observed fraction of nucleotide changes that are removed by purifying selection. In addition, a comparison of selective coefficients between drosophilids and hominids revealed significantly higher constraints in drosophilids, which can be attributed to the difference in long-term effective population size between these two groups of species. This difference is particularly apparent at the independently evolving (unpaired) sites.