Detection of Ghost Introgression Requires Exploiting Topological and Branch Length Information.

In recent years, the study of hybridization and introgression has made significant progress, with ghost introgression-the transfer of genetic material from extinct or unsampled lineages to extant species-emerging as a key area for research. Accurately identifying ghost introgression, however, presents a challenge. To address this issue, we focused on simple cases involving 3 species with a known phylogenetic tree. Using mathematical analyses and simulations, we evaluated the performance of popular phylogenetic methods, including HyDe and PhyloNet/MPL, and the full-likelihood method, Bayesian Phylogenetics and Phylogeography (BPP), in detecting ghost introgression. Our findings suggest that heuristic approaches relying on site-pattern counts or gene-tree topologies struggle to differentiate ghost introgression from introgression between sampled non-sister species, frequently leading to incorrect identification of donor and recipient species. The full-likelihood method BPP uses multilocus sequence alignments directly-hence taking into account both gene-tree topologies and branch lengths, by contrast, is capable of detecting ghost introgression in phylogenomic datasets. We analyzed a real-world phylogenomic dataset of 14 species of Jaltomata (Solanaceae) to showcase the potential of full-likelihood methods for accurate inference of introgression.

Deep Learning and Likelihood Approaches for Viral Phylogeography Converge on the Same Answers Whether the Inference Model Is Right or Wrong.

Analysis of phylogenetic trees has become an essential tool in epidemiology. Likelihood-based methods fit models to phylogenies to draw inferences about the phylodynamics and history of viral transmission. However, these methods are often computationally expensive, which limits the complexity and realism of phylodynamic models and makes them ill-suited for informing policy decisions in real-time during rapidly developing outbreaks. Likelihood-free methods using deep learning are pushing the boundaries of inference beyond these constraints. In this paper, we extend, compare, and contrast a recently developed deep learning method for likelihood-free inference from trees. We trained multiple deep neural networks using phylogenies from simulated outbreaks that spread among 5 locations and found they achieve close to the same levels of accuracy as Bayesian inference under the true simulation model. We compared robustness to model misspecification of a trained neural network to that of a Bayesian method. We found that both models had comparable performance, converging on similar biases. We also implemented a method of uncertainty quantification called conformalized quantile regression that we demonstrate has similar patterns of sensitivity to model misspecification as Bayesian highest posterior density (HPD) and greatly overlap with HPDs, but have lower precision (more conservative). Finally, we trained and tested a neural network against phylogeographic data from a recent study of the SARS-Cov-2 pandemic in Europe and obtained similar estimates of region-specific epidemiological parameters and the location of the common ancestor in Europe. Along with being as accurate and robust as likelihood-based methods, our trained neural networks are on average over 3 orders of magnitude faster after training. Our results support the notion that neural networks can be trained with simulated data to accurately mimic the good and bad statistical properties of the likelihood functions of generative phylogenetic models.

Accounting for spatial sampling patterns in Bayesian phylogeography.

Statistical phylogeography provides useful tools to characterize and quantify the spread of organisms during the course of evolution. Analyzing georeferenced genetic data often relies on the assumption that samples are preferentially collected in densely populated areas of the habitat. Deviation from this assumption negatively impacts the inference of the spatial and demographic dynamics. This issue is pervasive in phylogeography. It affects analyses that approximate the habitat as a set of discrete demes as well as those that treat it as a continuum. The present study introduces a Bayesian modeling approach that explicitly accommodates for spatial sampling strategies. An original inference technique, based on recent advances in statistical computing, is then described that is most suited to modeling data where sequences are preferentially collected at certain locations, independently of the outcome of the evolutionary process. The analysis of georeferenced genetic sequences from the West Nile virus in North America along with simulated data shows how assumptions about spatial sampling may impact our understanding of the forces shaping biodiversity across time and space.

Phylogeographic analysis of Pseudogymnoascus destructans partitivirus-pa explains the spread dynamics of white-nose syndrome in North America.

Understanding the dynamics of white-nose syndrome spread in time and space is an important component for the disease epidemiology and control. We reported earlier that a novel partitivirus, Pseudogymnoascus destructans partitivirus-pa, had infected the North American isolates of Pseudogymnoascus destructans, the fungal pathogen that causes white-nose syndrome in bats. We showed that the diversity of the viral coat protein sequences is correlated to their geographical origin. Here we hypothesize that the geographical adaptation of the virus could be used as a proxy to characterize the spread of white-nose syndrome. We used over 100 virus isolates from diverse locations in North America and applied the phylogeographic analysis tool BEAST to characterize the spread of the disease. The strict clock phylogeographic analysis under the coalescent model in BEAST showed a patchy spread pattern of white-nose syndrome driven from a few source locations including Connecticut, New York, West Virginia, and Kentucky. The source states had significant support in the maximum clade credibility tree and Bayesian stochastic search variable selection analysis. Although the geographic origin of the virus is not definite, it is likely the virus infected the fungus prior to the spread of white-nose syndrome in North America. We also inferred from the BEAST analysis that the recent long-distance spread of the fungus to Washington had its root in Kentucky, likely from the Mammoth cave area and most probably mediated by a human. The time to the most recent common ancestor of the virus is estimated somewhere between the late 1990s to early 2000s. We found the mean substitution rate of 2 X 10-3 substitutions per site per year for the virus which is higher than expected given the persistent lifestyle of the virus, and the stamping-machine mode of replication. Our approach of using the virus as a proxy to understand the spread of white-nose syndrome could be an important tool for the study and management of other infectious diseases.

Disproportionate extinction of South American mammals drove the asymmetry of the Great American Biotic Interchange.

The interchange between the previously disconnected faunas of North and South America was a massive experiment in biological invasion. A major gap in our understanding of this invasion is why there was a drastic increase in the proportion of mammals of North American origin found in South America. Four nonmutually exclusive mechanisms may explain this asymmetry: 1) Higher dispersal rate of North American mammals toward the south, 2) higher origination of North American immigrants in South America, 3) higher extinction of mammals with South American origin, and 4) similar dispersal rate but a larger pool of native taxa in North versus South America. We test among these mechanisms by analyzing ∼20,000 fossil occurrences with Bayesian methods to infer dispersal and diversification rates and taxonomic selectivity of immigrants. We find no differences in the dispersal and origination rates of immigrants. In contrast, native South American mammals show higher extinction. We also find that two clades with North American origin (Carnivora and Artiodactyla) had significantly more immigrants in South America than other clades. Altogether, the asymmetry of the interchange was not due to higher origination of immigrants in South America as previously suggested, but resulted from higher extinction of native taxa in southern South America. These results from one of the greatest biological invasions highlight how biogeographic processes and biotic interactions can shape continental diversity.

Relax, Keep Walking - A Practical Guide to Continuous Phylogeographic Inference with BEAST.

Spatially explicit phylogeographic analyses can be performed with an inference framework that employs relaxed random walks to reconstruct phylogenetic dispersal histories in continuous space. This core model was first implemented 10 years ago and has opened up new opportunities in the field of phylodynamics, allowing researchers to map and analyze the spatial dissemination of rapidly evolving pathogens. We here provide a detailed and step-by-step guide on how to set up, run, and interpret continuous phylogeographic analyses using the programs BEAUti, BEAST, Tracer, and TreeAnnotator.

Sampling bias and model choice in continuous phylogeography: Getting lost on a random walk.

Phylogeographic inference allows reconstruction of past geographical spread of pathogens or living organisms by integrating genetic and geographic data. A popular model in continuous phylogeography-with location data provided in the form of latitude and longitude coordinates-describes spread as a Brownian motion (Brownian Motion Phylogeography, BMP) in continuous space and time, akin to similar models of continuous trait evolution. Here, we show that reconstructions using this model can be strongly affected by sampling biases, such as the lack of sampling from certain areas. As an attempt to reduce the effects of sampling bias on BMP, we consider the addition of sequence-free samples from under-sampled areas. While this approach alleviates the effects of sampling bias, in most scenarios this will not be a viable option due to the need for prior knowledge of an outbreak's spatial distribution. We therefore consider an alternative model, the spatial Λ-Fleming-Viot process (ΛFV), which has recently gained popularity in population genetics. Despite the ΛFV's robustness to sampling biases, we find that the different assumptions of the ΛFV and BMP models result in different applicabilities, with the ΛFV being more appropriate for scenarios of endemic spread, and BMP being more appropriate for recent outbreaks or colonizations.

A Southeast Asian origin for present-day non-African human Y chromosomes.

The genomes of present-day humans outside Africa originated almost entirely from a single out-migration ~ 50,000-70,000 years ago, followed by mixture with Neanderthals contributing ~ 2% to all non-Africans. However, the details of this initial migration remain poorly understood because no ancient DNA analyses are available from this key time period, and interpretation of present-day autosomal data is complicated due to subsequent population movements/reshaping. One locus, however, does retain male-specific information from this early period: the Y chromosome, where a detailed calibrated phylogeny has been constructed. Three present-day Y lineages were carried by the initial migration: the rare haplogroup D, the moderately rare C, and the very common FT lineage which now dominates most non-African populations. Here, we show that phylogenetic analyses of haplogroup C, D and FT sequences, including very rare deep-rooting lineages, together with phylogeographic analyses of ancient and present-day non-African Y chromosomes, all point to East/Southeast Asia as the origin 50,000-55,000 years ago of all known surviving non-African male lineages (apart from recent migrants). This observation contrasts with the expectation of a West Eurasian origin predicted by a simple model of expansion from a source near Africa, and can be interpreted as resulting from extensive genetic drift in the initial population or replacement of early western Y lineages from the east, thus informing and constraining models of the initial expansion.

Phylogeography and Antigenic Diversity of Low-Pathogenic Avian Influenza H13 and H16 Viruses.

Low-pathogenic avian influenza viruses (LPAIVs) are genetically highly variable and have diversified into multiple evolutionary lineages that are primarily associated with wild-bird reservoirs. Antigenic variation has been described for mammalian influenza viruses and for highly pathogenic avian influenza viruses that circulate in poultry, but much less is known about antigenic variation of LPAIVs. In this study, we focused on H13 and H16 LPAIVs that circulate globally in gulls. We investigated the evolutionary history and intercontinental gene flow based on the hemagglutinin (HA) gene and used representative viruses from genetically distinct lineages to determine their antigenic properties by hemagglutination inhibition assays. For H13, at least three distinct genetic clades were evident, while for H16, at least two distinct genetic clades were evident. Twenty and ten events of intercontinental gene flow were identified for H13 and H16 viruses, respectively. At least two antigenic variants of H13 and at least one antigenic variant of H16 were identified. Amino acid positions in the HA protein that may be involved in the antigenic variation were inferred, and some of the positions were located near the receptor binding site of the HA protein, as they are in the HA protein of mammalian influenza A viruses. These findings suggest independent circulation of H13 and H16 subtypes in gull populations, as antigenic patterns do not overlap, and they contribute to the understanding of the genetic and antigenic variation of LPAIVs naturally circulating in wild birds.IMPORTANCE Wild birds play a major role in the epidemiology of low-pathogenic avian influenza viruses (LPAIVs), which are occasionally transmitted-directly or indirectly-from them to other species, including domestic animals, wild mammals, and humans, where they can cause subclinical to fatal disease. Despite a multitude of genetic studies, the antigenic variation of LPAIVs in wild birds is poorly understood. Here, we investigated the evolutionary history, intercontinental gene flow, and antigenic variation among H13 and H16 LPAIVs. The circulation of subtypes H13 and H16 seems to be maintained by a narrower host range, in particular gulls, than the majority of LPAIV subtypes and may therefore serve as a model for evolution and epidemiology of H1 to H12 LPAIVs in wild birds. The findings suggest that H13 and H16 LPAIVs circulate independently of each other and emphasize the need to investigate within-clade antigenic variation of LPAIVs in wild birds.

Population structure and gene flow of the syntopic turtles Emys and Mauremys from coastal and inland regions of Anatolia (Turkey): results from mitochondrial and microsatellite data.

Revealing the genetic basis of the existence of different species living together in different geographic regions provides clarification of this phylogeographic differentiation. In this study, we investigated the population genetics and evaluated the level of genetic variation of inland and coastal populations of Mauremys and Emys in Turkey. Tissue samples of 196 terrapins were studied which were collected from syntopic coastal (Gölbent-Söke/Aydin; M. rivulata and E. orbicularis) and inland populations (Bahçesaray/Aksaray; M. caspica and E. orbicularis). DNA was isolated using the InnuPREP DNA Mini Kit. Mitochondrial DNA sequences and allelic variation at 13 microsatellite loci for Mauremys and 12 microsatellite loci for Emys were examined.  Three haplotypes were found for Emys orbicularis (Im, Ip and Iw) collected from the coastal region and two haplotypes for Emys orbicularis (Ig and Im) collected from inland. Two haplotypes were identified for M. caspica (Cmt8 and Cmt9) and three haplotypes were identified for M. rivulata (Rmt3, Rmt24 and Rmt26). Using microsatellites and the software STRUCTURE the most probable value for K was revealed as two 2 for both species. The FST value between M. rivulata and M. caspica was 0.39, and between the coastal and inland populations of E. orbicularis 0.09. It can be concluded that Emys populations tend to evolve by somehow preserving the allelic richness they have and Mauremys populations continue to differentiate so that new species emerge in the evolutionary process to reach the ideal allelic structure.

Contrasting responses of functional diversity to major losses in taxonomic diversity.

Taxonomic diversity of benthic marine invertebrate shelf species declines at present by nearly an order of magnitude from the tropics to the poles in each hemisphere along the latitudinal diversity gradient (LDG), most steeply along the western Pacific where shallow-sea diversity is at its tropical maximum. In the Bivalvia, a model system for macroevolution and macroecology, this taxonomic trend is accompanied by a decline in the number of functional groups and an increase in the evenness of taxa distributed among those groups, with maximum functional evenness (FE) in polar waters of both hemispheres. In contrast, analyses of this model system across the two era-defining events of the Phanerozoic, the Permian-Triassic and Cretaceous-Paleogene mass extinctions, show only minor declines in functional richness despite high extinction intensities, resulting in a rise in FE owing to the persistence of functional groups. We hypothesize that the spatial decline of taxonomic diversity and increase in FE along the present-day LDG primarily reflect diversity-dependent factors, whereas retention of almost all functional groups through the two mass extinctions suggests the operation of diversity-independent factors. Comparative analyses of different aspects of biodiversity thus reveal strongly contrasting biological consequences of similarly severe declines in taxonomic diversity and can help predict the consequences for functional diversity among different drivers of past, present, and future biodiversity loss.

Paleocene Ipomoea (Convolvulaceae) from India with implications for an East Gondwana origin of Convolvulaceae.

The morning glory family, Convolvulaceae, is globally important in medicine and food crops. The family has worldwide distribution in a variety of habitats; however, its fossil record is very poorly documented. The current fossil record suggests an origin in North America, which is in contrast to molecular data that indicate an East Gondwana origin. We report Ipomoea leaves from the late Paleocene (Thanetian; 58.7-55.8 million years ago) of India, which was a part of East Gondwana during this time. This is the earliest fossil record for both the family Convolvulaceae and the order Solanales. This suggests that the sister families Convolvulaceae and Solanaceae diverged before the Eocene in Gondwana-derived continents. The evidence presented here supports the conclusion from molecular phylogenetic analysis of an East Gondwana origin of Convolvulaceae.

Insular Microbiogeography: Three Pathogens as Exemplars.

Traditional taxonomy in biology assumes that life is organized in a simple tree. Attempts to classify microorganisms in this way in the genomics era led microbiologists to look for finite sets of 'core' genes that uniquely group taxa as clades in the tree. However, the diversity revealed by large-scale whole genome sequencing is calling into question the long-held model of a hierarchical tree of life, which leads to questioning of the definition of a species. Large-scale studies of microbial genome diversity reveal that the cumulative number of new genes discovered increases with the number of genomes studied as a power law and subsequently leads to the lack of evidence for a unique core genome within closely related organisms. Sampling 'enough' new genomes leads to the discovery of a replacement or alternative to any gene. This power law behaviour points to an underlying self-organizing critical process that may be guided by mutation and niche selection. Microbes in any particular niche exist within a local web of organism interdependence known as the microbiome. The same mechanism that underpins the macro-ecological scaling first observed by MacArthur and Wilson also applies to microbial communities. Recent metagenomic studies of a food microbiome demonstrate the diverse distribution of community members, but also genotypes for a single species within a more complex community. Collectively, these results suggest that traditional taxonomic classification of bacteria could be replaced with a quasispecies model. This model is commonly accepted in virology and better describes the diversity and dynamic exchange of genes that also hold true for bacteria. This model will enable microbiologists to conduct population-scale studies to describe microbial behaviour, as opposed to a single isolate as a representative.

Genome-wide profiles indicate wolf population connectivity within the eastern Carpathian Mountains.

The Carpathian Mountains provide critical wildlife habitat in central Europe, and previous genome-wide studies have found western Carpathian Mountain wolves (Canis lupus) to be a separate population. Whereas differentiation to the north may be explained by a lowland-mountain transition and habitat fragmentation, the eastern Carpathian Mountains extending through Romania appear to offer continuous wildlife habitat southward. Our objective was to assess gene flow patterns and population connectivity among wolves in Romania, western Ukraine, and the Republic of Moldova. We sought to determine if the Carpathian Mountain region is best described by a north-south gradient in genetic profiles, or whether Romanian wolves show population structure with northern individuals clustering with western Ukraine. We genotyped 48 individuals with 170 000 single nucleotide polymorphism markers, and successful profiles from Romania (n = 27) and Moldova (n = 2) were merged with existing data from western Ukraine (n = 10). Expected heterozygosity was 0.234 (SE 0.001) for Romania and 0.229 (SE 0.001) for western Ukraine, whereas observed heterozygosity values were 0.230 (SE 0.001) versus 0.231 (SE 0.001). Population structure analyses with a maximum likelihood method supported K = 1 population, followed by K = 2 where Romania formed one cluster, and western Ukraine and Moldova formed another. Principal component analysis results were broadly consistent with K = 2. Pairwise FST between western Ukraine and Romania was 0.042 (p = 0.001). Our findings indicated weak population differentiation, and future research may clarify whether the spatial distribution of genetic diversity in the region is associated with environmental and ecological factors such as terrain ruggedness and the distribution of prey species.

On the origin of European sheep as revealed by the diversity of the Balkan breeds and by optimizing population-genetic analysis tools.

BACKGROUND: In the Neolithic, domestic sheep migrated into Europe and subsequently spread in westerly and northwesterly directions. Reconstruction of these migrations and subsequent genetic events requires a more detailed characterization of the current phylogeographic differentiation. RESULTS: We collected 50 K single nucleotide polymorphism (SNP) profiles of Balkan sheep that are currently found near the major Neolithic point of entry into Europe, and combined these data with published genotypes from southwest-Asian, Mediterranean, central-European and north-European sheep and from Asian and European mouflons. We detected clines, ancestral components and admixture by using variants of common analysis tools: geography-informative supervised principal component analysis (PCA), breed-specific admixture analysis, across-breed [Formula: see text] profiles and phylogenetic analysis of regional pools of breeds. The regional Balkan sheep populations exhibit considerable genetic overlap, but are clearly distinct from the breeds in surrounding regions. The Asian mouflon did not influence the differentiation of the European domestic sheep and is only distantly related to present-day sheep, including those from Iran where the mouflons were sampled. We demonstrate the occurrence, from southeast to northwest Europe, of a continuously increasing ancestral component of up to 20% contributed by the European mouflon, which is assumed to descend from the original Neolithic domesticates. The overall patterns indicate that the Balkan region and Italy served as post-domestication migration hubs, from which wool sheep reached Spain and north Italy with subsequent migrations northwards. The documented dispersal of Tarentine wool sheep during the Roman period may have been part of this process. Our results also reproduce the documented 18th century admixture of Spanish Merino sheep into several central-European breeds. CONCLUSIONS: Our results contribute to a better understanding of the events that have created the present diversity pattern, which is relevant for the management of the genetic resources represented by the European sheep population.

Effect of river size on Amazonian primate community structure: A biogeographic analysis using updated taxonomic assessments.

The mechanisms that underlie the diversification of Neotropical primates remain contested. One mechanism that has found support is the riverine barrier hypothesis (RBH), which postulates that large rivers impede gene flow between populations on opposite riverbanks and promote allopatric speciation. Ayres and Clutton-Brock (1992) demonstrated that larger Amazonian rivers acted as barriers, delineating the distribution limits of primate species. However, profound changes in taxonomy and species concepts have led to the proliferation of Neotropical primate taxa, which may have reduced support for their results. Using the most recent taxonomic assessments and distribution maps, we tested the effect of increasing river size on the similarity of opposite riverbank primate communities in the Amazon. First, we conducted a literature review of primate taxonomy and developed a comprehensive spatial database, then applied geographical information system to query mapped primate ranges against the riverine geography of the Amazon watershed to produce a similarity index for opposite riverbank communities. Finally, we ran models to test how measures of river size predicted levels of similarity. We found that, almost without exception, similarity scores were lower than scores from Ayres and Clutton-Brock (1992) for the same rivers. Our model showed a significant negative relationship between streamflow and similarity in all tests, and found river width significant for the segmented Amazon, but not for multiple Amazon watershed rivers. Our results support the RBH insofar as they provide evidence for the prediction that rivers with higher streamflow act as more substantial barriers to dispersal, and accordingly exhibit greater variation in community composition between riverbanks.

Spatial and Temporal Scales of Range Expansion in Wild Phaseolus vulgaris.

The wild progenitor of common-bean has an exceptionally large distribution from northern Mexico to northwestern Argentina, unusual among crop wild progenitors. This research sought to document major events of range expansion that led to this distribution and associated environmental changes. Through the use of genotyping-by-sequencing (∼20,000 SNPs) and geographic information systems applied to a sample of 246 accessions of wild Phaseolus vulgaris, including 157 genotypes of the Mesoamerican, 77 of the southern Andean, and 12 of the Northern Peru-Ecuador gene pools, we identified five geographically distinct subpopulations. Three of these subpopulations belong to the Mesoamerican gene pool (Northern and Central Mexico, Oaxaca, and Southern Mexico, Central America and northern South America) and one each to the Northern Peru-Ecuador (PhI) and the southern Andean gene pools. The five subpopulations were distributed in different floristic provinces of the Neotropical seasonally dry forest and showed distinct distributions for temperature and rainfall resulting in decreased local potential evapotranspiration (PhI and southern Andes groups) compared with the two Mexican groups. Three of these subpopulations represent long-distance dispersal events from Mesoamerica into Northern Peru-Ecuador, southern Andes, and Central America and Colombia, in chronological order. Of particular note is that the dispersal to Northern Peru-Ecuador markedly predates the dispersal to the southern Andes (∼400 vs. ∼100 ky), consistent with the ancestral nature of the phaseolin seed protein and chloroplast sequences observed in the PhI group. Seed dispersal in common bean can be, therefore, described at different spatial and temporal scales, from localized, annual seed shattering to long-distance, evolutionarily rare migration.

Out of Southern East Asia of the Brown Rat Revealed by Large-Scale Genome Sequencing.

The geographic origin and migration of the brown rat (Rattus norvegicus) remain subjects of considerable debate. In this study, we sequenced whole genomes of 110 wild brown rats with a diverse world-wide representation. We reveal that brown rats migrated out of southern East Asia, rather than northern Asia as formerly suggested, into the Middle East and then to Europe and Africa, thousands of years ago. Comparison of genomes from different geographical populations reveals that many genes involved in the immune system experienced positive selection in the wild brown rat.

Global distribution of mating types shows limited opportunities for mating across populations of fungi causing boxwood blight disease.

Boxwood blight is a disease threat to natural and managed landscapes worldwide. To determine mating potential of the fungi responsible for the disease, Calonectria pseudonaviculata and C. henricotiae, we characterized their mating-type (MAT) loci. Genomes of C. henricotiae, C. pseudonaviculata and two other Calonectria species (C. leucothoes, C. naviculata) were sequenced and used to design PCR tests for mating-type from 268 isolates collected from four continents. All four Calonectria species have a MAT locus that is structurally consistent with the organization found in heterothallic ascomycetes, with just one idiomorph per individual isolate. Mating type was subdivided by species: all C. henricotiae isolates possessed the MAT1-1 idiomorph, whereas all C. pseudonaviculata isolates possessed the MAT1-2 idiomorph. To determine the potential for divergence at the MAT1 locus to present a barrier to interspecific hybridization, evolutionary analysis was conducted. Phylogenomic estimates showed that C. henricotiae and C. pseudonaviculata diverged approximately 2.1 Mya. However, syntenic comparisons, phylogenetic analyses, and estimates of nucleotide divergence across the MAT1 locus and proximal genes identified minimal divergence in this region of the genome. These results show that in North America and parts of Europe, where only C. pseudonaviculata resides, mating is constrained by the absence of MAT1-1. In regions of Europe where C. henricotiae and C. pseudonaviculata currently share the same host and geographic range, it remains to be determined whether or not these two recently diverged species are able to overcome species barriers to mate.

Deep neural networks and distant supervision for geographic location mention extraction.

Motivation: Virus phylogeographers rely on DNA sequences of viruses and the locations of the infected hosts found in public sequence databases like GenBank for modeling virus spread. However, the locations in GenBank records are often only at the country or state level, and may require phylogeographers to scan the journal articles associated with the records to identify more localized geographic areas. To automate this process, we present a named entity recognizer (NER) for detecting locations in biomedical literature. We built the NER using a deep feedforward neural network to determine whether a given token is a toponym or not. To overcome the limited human annotated data available for training, we use distant supervision techniques to generate additional samples to train our NER. Results: Our NER achieves an F1-score of 0.910 and significantly outperforms the previous state-of-the-art system. Using the additional data generated through distant supervision further boosts the performance of the NER achieving an F1-score of 0.927. The NER presented in this research improves over previous systems significantly. Our experiments also demonstrate the NER's capability to embed external features to further boost the system's performance. We believe that the same methodology can be applied for recognizing similar biomedical entities in scientific literature.