Many-core algorithms for high-dimensional gradients on phylogenetic trees.

MOTIVATION: Advancements in high-throughput genomic sequencing are delivering genomic pathogen data at an unprecedented rate, positioning statistical phylogenetics as a critical tool to monitor infectious diseases globally. This rapid growth spurs the need for efficient inference techniques, such as Hamiltonian Monte Carlo (HMC) in a Bayesian framework, to estimate parameters of these phylogenetic models where the dimensions of the parameters increase with the number of sequences N. HMC requires repeated calculation of the gradient of the data log-likelihood with respect to (wrt) all branch-length-specific (BLS) parameters that traditionally takes O(N2) operations using the standard pruning algorithm. A recent study proposes an approach to calculate this gradient in O(N), enabling researchers to take advantage of gradient-based samplers such as HMC. The CPU implementation of this approach makes the calculation of the gradient computationally tractable for nucleotide-based models but falls short in performance for larger state-space size models, such as Markov-modulated and codon models. Here, we describe novel massively parallel algorithms to calculate the gradient of the log-likelihood wrt all BLS parameters that take advantage of graphics processing units (GPUs) and result in many fold higher speedups over previous CPU implementations. RESULTS: We benchmark these GPU algorithms on three computing systems using three evolutionary inference examples exploring complete genomes from 997 dengue viruses, 62 carnivore mitochondria and 49 yeasts, and observe a >128-fold speedup over the CPU implementation for codon-based models and >8-fold speedup for nucleotide-based models. As a practical demonstration, we also estimate the timing of the first introduction of West Nile virus into the continental Unites States under a codon model with a relaxed molecular clock from 104 full viral genomes, an inference task previously intractable. AVAILABILITY AND IMPLEMENTATION: We provide an implementation of our GPU algorithms in BEAGLE v4.0.0 (https://github.com/beagle-dev/beagle-lib), an open-source library for statistical phylogenetics that enables parallel calculations on multi-core CPUs and GPUs. We employ a BEAGLE-implementation using the Bayesian phylogenetics framework BEAST (https://github.com/beast-dev/beast-mcmc).

Genomic surveillance reveals dynamic shifts in the connectivity of COVID-19 epidemics.

The maturation of genomic surveillance in the past decade has enabled tracking of the emergence and spread of epidemics at an unprecedented level. During the COVID-19 pandemic, for example, genomic data revealed that local epidemics varied considerably in the frequency of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) lineage importation and persistence, likely due to a combination of COVID-19 restrictions and changing connectivity. Here, we show that local COVID-19 epidemics are driven by regional transmission, including across international boundaries, but can become increasingly connected to distant locations following the relaxation of public health interventions. By integrating genomic, mobility, and epidemiological data, we find abundant transmission occurring between both adjacent and distant locations, supported by dynamic mobility patterns. We find that changing connectivity significantly influences local COVID-19 incidence. Our findings demonstrate a complex meaning of "local" when investigating connected epidemics and emphasize the importance of collaborative interventions for pandemic prevention and mitigation.

Genetic tracing of market wildlife and viruses at the epicenter of the COVID-19 pandemic.

Zoonotic spillovers of viruses have occurred through the animal trade worldwide. The start of the COVID-19 pandemic was traced epidemiologically to the Huanan Wholesale Seafood Market, the site with the most reported wildlife vendors in the city of Wuhan, China. Here, we analyze publicly available qPCR and sequencing data from environmental samples collected in the Huanan market in early 2020. We demonstrate that the SARS-CoV-2 genetic diversity linked to this market is consistent with market emergence, and find increased SARS-CoV-2 positivity near and within a particular wildlife stall. We identify wildlife DNA in all SARS-CoV-2 positive samples from this stall. This includes species such as civets, bamboo rats, porcupines, hedgehogs, and one species, raccoon dogs, known to be capable of SARS-CoV-2 transmission. We also detect other animal viruses that infect raccoon dogs, civets, and bamboo rats. Combining metagenomic and phylogenetic approaches, we recover genotypes of market animals and compare them to those from other markets. This analysis provides the genetic basis for a short list of potential intermediate hosts of SARS-CoV-2 to prioritize for retrospective serological testing and viral sampling.

Automatic Differentiation is no Panacea for Phylogenetic Gradient Computation.

Gradients of probabilistic model likelihoods with respect to their parameters are essential for modern computational statistics and machine learning. These calculations are readily available for arbitrary models via "automatic differentiation" implemented in general-purpose machine-learning libraries such as TensorFlow and PyTorch. Although these libraries are highly optimized, it is not clear if their general-purpose nature will limit their algorithmic complexity or implementation speed for the phylogenetic case compared to phylogenetics-specific code. In this paper, we compare six gradient implementations of the phylogenetic likelihood functions, in isolation and also as part of a variational inference procedure. We find that although automatic differentiation can scale approximately linearly in tree size, it is much slower than the carefully implemented gradient calculation for tree likelihood and ratio transformation operations. We conclude that a mixed approach combining phylogenetic libraries with machine learning libraries will provide the optimal combination of speed and model flexibility moving forward.

Many-core algorithms for high-dimensional gradients on phylogenetic trees.

The rapid growth in genomic pathogen data spurs the need for efficient inference techniques, such as Hamiltonian Monte Carlo (HMC) in a Bayesian framework, to estimate parameters of these phylogenetic models where the dimensions of the parameters increase with the number of sequences $N$. HMC requires repeated calculation of the gradient of the data log-likelihood with respect to (wrt) all branch-length-specific (BLS) parameters that traditionally takes $\mathcal{O}(N^2)$ operations using the standard pruning algorithm. A recent study proposes an approach to calculate this gradient in $\mathcal{O}(N)$, enabling researchers to take advantage of gradient-based samplers such as HMC. The CPU implementation of this approach makes the calculation of the gradient computationally tractable for nucleotide-based models but falls short in performance for larger state-space size models, such as codon models. Here, we describe novel massively parallel algorithms to calculate the gradient of the log-likelihood wrt all BLS parameters that take advantage of graphics processing units (GPUs) and result in many fold higher speedups over previous CPU implementations. We benchmark these GPU algorithms on three computing systems using three evolutionary inference examples: carnivores, dengue and yeast, and observe a greater than 128-fold speedup over the CPU implementation for codon-based models and greater than 8-fold speedup for nucleotide-based models. As a practical demonstration, we also estimate the timing of the first introduction of West Nile virus into the continental Unites States under a codon model with a relaxed molecular clock from 104 full viral genomes, an inference task previously intractable. We provide an implementation of our GPU algorithms in BEAGLE v4.0.0, an open source library for statistical phylogenetics that enables parallel calculations on multi-core CPUs and GPUs.

Outbreak.info Research Library: a standardized, searchable platform to discover and explore COVID-19 resources.

Outbreak.info Research Library is a standardized, searchable interface of coronavirus disease 2019 (COVID-19) and severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) publications, clinical trials, datasets, protocols and other resources, built with a reusable framework. We developed a rigorous schema to enforce consistency across different sources and resource types and linked related resources. Researchers can quickly search the latest research across data repositories, regardless of resource type or repository location, via a search interface, public application programming interface (API) and R package.

Outbreak.info genomic reports: scalable and dynamic surveillance of SARS-CoV-2 variants and mutations.

In response to the emergence of SARS-CoV-2 variants of concern, the global scientific community, through unprecedented effort, has sequenced and shared over 11 million genomes through GISAID, as of May 2022. This extraordinarily high sampling rate provides a unique opportunity to track the evolution of the virus in near real-time. Here, we present outbreak.info , a platform that currently tracks over 40 million combinations of Pango lineages and individual mutations, across over 7,000 locations, to provide insights for researchers, public health officials and the general public. We describe the interpretable visualizations available in our web application, the pipelines that enable the scalable ingestion of heterogeneous sources of SARS-CoV-2 variant data and the server infrastructure that enables widespread data dissemination via a high-performance API that can be accessed using an R package. We show how outbreak.info can be used for genomic surveillance and as a hypothesis-generation tool to understand the ongoing pandemic at varying geographic and temporal scales.

Predicting the evolution of the Lassa virus endemic area and population at risk over the next decades.

Lassa fever is a severe viral hemorrhagic fever caused by a zoonotic virus that repeatedly spills over to humans from its rodent reservoirs. It is currently not known how climate and land use changes could affect the endemic area of this virus, currently limited to parts of West Africa. By exploring the environmental data associated with virus occurrence using ecological niche modelling, we show how temperature, precipitation and the presence of pastures determine ecological suitability for virus circulation. Based on projections of climate, land use, and population changes, we find that regions in Central and East Africa will likely become suitable for Lassa virus over the next decades and estimate that the total population living in ecological conditions that are suitable for Lassa virus circulation may drastically increase by 2070. By analysing geotagged viral genomes using spatially-explicit phylogeography and simulating virus dispersal, we find that in the event of Lassa virus being introduced into a new suitable region, its spread might remain spatially limited over the first decades.

Regional connectivity drove bidirectional transmission of SARS-CoV-2 in the Middle East during travel restrictions.

Regional connectivity and land travel have been identified as important drivers of SARS-CoV-2 transmission. However, the generalizability of this finding is understudied outside of well-sampled, highly connected regions. In this study, we investigated the relative contributions of regional and intercontinental connectivity to the source-sink dynamics of SARS-CoV-2 for Jordan and the Middle East. By integrating genomic, epidemiological and travel data we show that the source of introductions into Jordan was dynamic across 2020, shifting from intercontinental seeding in the early pandemic to more regional seeding for the travel restrictions period. We show that land travel, particularly freight transport, drove introduction risk during the travel restrictions period. High regional connectivity and land travel also drove Jordan's export risk. Our findings emphasize regional connectedness and land travel as drivers of transmission in the Middle East.

Outbreak.info genomic reports: scalable and dynamic surveillance of SARS-CoV-2 variants and mutations.

The emergence of SARS-CoV-2 variants of concern has prompted the need for near real-time genomic surveillance to inform public health interventions. In response to this need, the global scientific community, through unprecedented effort, has sequenced and shared over 11 million genomes through GISAID, as of May 2022. This extraordinarily high sampling rate provides a unique opportunity to track the evolution of the virus in near real-time. Here, we present outbreak.info, a platform that currently tracks over 40 million combinations of PANGO lineages and individual mutations, across over 7,000 locations, to provide insights for researchers, public health officials, and the general public. We describe the interpretable and opinionated visualizations in the variant and location focussed reports available in our web application, the pipelines that enable the scalable ingestion of heterogeneous sources of SARS-CoV-2 variant data, and the server infrastructure that enables widespread data dissemination via a high performance API that can be accessed using an R package. We present a case study that illustrates how outbreak.info can be used for genomic surveillance and as a hypothesis generation tool to understand the ongoing pandemic at varying geographic and temporal scales. With an emphasis on scalability, interactivity, interpretability, and reusability, outbreak.info provides a template to enable genomic surveillance at a global and localized scale.

The molecular epidemiology of multiple zoonotic origins of SARS-CoV-2.

Understanding the circumstances that lead to pandemics is important for their prevention. We analyzed the genomic diversity of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) early in the coronavirus disease 2019 (COVID-19) pandemic. We show that SARS-CoV-2 genomic diversity before February 2020 likely comprised only two distinct viral lineages, denoted "A" and "B." Phylodynamic rooting methods, coupled with epidemic simulations, reveal that these lineages were the result of at least two separate cross-species transmission events into humans. The first zoonotic transmission likely involved lineage B viruses around 18 November 2019 (23 October to 8 December), and the separate introduction of lineage A likely occurred within weeks of this event. These findings indicate that it is unlikely that SARS-CoV-2 circulated widely in humans before November 2019 and define the narrow window between when SARS-CoV-2 first jumped into humans and when the first cases of COVID-19 were reported. As with other coronaviruses, SARS-CoV-2 emergence likely resulted from multiple zoonotic events.

Outbreak.info Research Library: A standardized, searchable platform to discover and explore COVID-19 resources.

To combat the ongoing COVID-19 pandemic, scientists have been conducting research at breakneck speeds, producing over 52,000 peer-reviewed articles within the first year. To address the challenge in tracking the vast amount of new research located in separate repositories, we developed outbreak.info Research Library, a standardized, searchable interface of COVID-19 and SARS-CoV-2 resources. Unifying metadata from sixteen repositories, we assembled a collection of over 350,000 publications, clinical trials, datasets, protocols, and other resources as of October 2022. We used a rigorous schema to enforce consistency across different sources and resource types and linked related resources. Researchers can quickly search the latest research across data repositories, regardless of resource type or repository location, via a search interface, public API, and R package. Finally, we discuss the challenges inherent in combining metadata from scattered and heterogeneous resources and provide recommendations to streamline this process to aid scientific research.

Emergence of an early SARS-CoV-2 epidemic in the United States.

The emergence of the COVID-19 epidemic in the United States (U.S.) went largely undetected due to inadequate testing. New Orleans experienced one of the earliest and fastest accelerating outbreaks, coinciding with Mardi Gras. To gain insight into the emergence of SARS-CoV-2 in the U.S. and how large-scale events accelerate transmission, we sequenced SARS-CoV-2 genomes during the first wave of the COVID-19 epidemic in Louisiana. We show that SARS-CoV-2 in Louisiana had limited diversity compared to other U.S. states and that one introduction of SARS-CoV-2 led to almost all of the early transmission in Louisiana. By analyzing mobility and genomic data, we show that SARS-CoV-2 was already present in New Orleans before Mardi Gras, and the festival dramatically accelerated transmission. Our study provides an understanding of how superspreading during large-scale events played a key role during the early outbreak in the U.S. and can greatly accelerate epidemics.

Ebola vaccine-induced protection in nonhuman primates correlates with antibody specificity and Fc-mediated effects.

Although substantial progress has been made with Ebola virus (EBOV) vaccine measures, the immune correlates of vaccine-mediated protection remain uncertain. Here, five mucosal vaccine vectors based on human and avian paramyxoviruses provided nonhuman primates with varying degrees of protection, despite expressing the same EBOV glycoprotein (GP) immunogen. Each vaccine produced antibody responses that differed in Fc-mediated functions and isotype composition, as well as in magnitude and coverage toward GP and its conformational and linear epitopes. Differences in the degree of protection and comprehensive characterization of the response afforded the opportunity to identify which features and functions were elevated in survivors and could therefore serve as vaccine correlates of protection. Pairwise network correlation analysis of 139 immune- and vaccine-related parameters was performed to demonstrate relationships with survival. Total GP-specific antibodies, as measured by biolayer interferometry, but not neutralizing IgG or IgA titers, correlated with survival. Fc-mediated functions and the amount of receptor binding domain antibodies were associated with improved survival outcomes, alluding to the protective mechanisms of these vaccines. Therefore, functional qualities of the antibody response, particularly Fc-mediated effects and GP specificity, rather than simply magnitude of the response, appear central to vaccine-induced protection against EBOV. The heterogeneity of the response profile between the vaccines indicates that each vaccine likely exhibits its own protective signature and the requirements for an efficacious EBOV vaccine are complex.

Emergence and rapid transmission of SARS-CoV-2 B.1.1.7 in the United States.

The highly transmissible B.1.1.7 variant of SARS-CoV-2, first identified in the United Kingdom, has gained a foothold across the world. Using S gene target failure (SGTF) and SARS-CoV-2 genomic sequencing, we investigated the prevalence and dynamics of this variant in the United States (US), tracking it back to its early emergence. We found that, while the fraction of B.1.1.7 varied by state, the variant increased at a logistic rate with a roughly weekly doubling rate and an increased transmission of 40%-50%. We revealed several independent introductions of B.1.1.7 into the US as early as late November 2020, with community transmission spreading it to most states within months. We show that the US is on a similar trajectory as other countries where B.1.1.7 became dominant, requiring immediate and decisive action to minimize COVID-19 morbidity and mortality.

Ebola Virus Transmission Initiated by Relapse of Systemic Ebola Virus Disease.

During the 2018-2020 Ebola virus disease (EVD) outbreak in North Kivu province in the Democratic Republic of Congo, EVD was diagnosed in a patient who had received the recombinant vesicular stomatitis virus-based vaccine expressing a ZEBOV glycoprotein (rVSV-ZEBOV) (Merck). His treatment included an Ebola virus (EBOV)-specific monoclonal antibody (mAb114), and he recovered within 14 days. However, 6 months later, he presented again with severe EVD-like illness and EBOV viremia, and he died. We initiated epidemiologic and genomic investigations that showed that the patient had had a relapse of acute EVD that led to a transmission chain resulting in 91 cases across six health zones over 4 months. (Funded by the Bill and Melinda Gates Foundation and others.).

Genomic epidemiology identifies emergence and rapid transmission of SARS-CoV-2 B.1.1.7 in the United States.

As of January of 2021, the highly transmissible B.1.1.7 variant of SARS-CoV-2, which was first identified in the United Kingdom (U.K.), has gained a strong foothold across the world. Because of the sudden and rapid rise of B.1.1.7, we investigated the prevalence and growth dynamics of this variant in the United States (U.S.), tracking it back to its early emergence and onward local transmission. We found that the RT-qPCR testing anomaly of S gene target failure (SGTF), first observed in the U.K., was a reliable proxy for B.1.1.7 detection. We sequenced 212 B.1.1.7 SARS-CoV-2 genomes collected from testing facilities in the U.S. from December 2020 to January 2021. We found that while the fraction of B.1.1.7 among SGTF samples varied by state, detection of the variant increased at a logistic rate similar to those observed elsewhere, with a doubling rate of a little over a week and an increased transmission rate of 35-45%. By performing time-aware Bayesian phylodynamic analyses, we revealed several independent introductions of B.1.1.7 into the U.S. as early as late November 2020, with onward community transmission enabling the variant to spread to at least 30 states as of January 2021. Our study shows that the U.S. is on a similar trajectory as other countries where B.1.1.7 rapidly became the dominant SARS-CoV-2 variant, requiring immediate and decisive action to minimize COVID-19 morbidity and mortality.

Emergence of an early SARS-CoV-2 epidemic in the United States.

The emergence of the early COVID-19 epidemic in the United States (U.S.) went largely undetected, due to a lack of adequate testing and mitigation efforts. The city of New Orleans, Louisiana experienced one of the earliest and fastest accelerating outbreaks, coinciding with the annual Mardi Gras festival, which went ahead without precautions. To gain insight into the emergence of SARS-CoV-2 in the U.S. and how large, crowded events may have accelerated early transmission, we sequenced SARS-CoV-2 genomes during the first wave of the COVID-19 epidemic in Louisiana. We show that SARS-CoV-2 in Louisiana initially had limited sequence diversity compared to other U.S. states, and that one successful introduction of SARS-CoV-2 led to almost all of the early SARS-CoV-2 transmission in Louisiana. By analyzing mobility and genomic data, we show that SARS-CoV-2 was already present in New Orleans before Mardi Gras and that the festival dramatically accelerated transmission, eventually leading to secondary localized COVID-19 epidemics throughout the Southern U.S.. Our study provides an understanding of how superspreading during large-scale events played a key role during the early outbreak in the U.S. and can greatly accelerate COVID-19 epidemics on a local and regional scale.

Epidemiological hypothesis testing using a phylogeographic and phylodynamic framework.

Computational analyses of pathogen genomes are increasingly used to unravel the dispersal history and transmission dynamics of epidemics. Here, we show how to go beyond historical reconstructions and use spatially-explicit phylogeographic and phylodynamic approaches to formally test epidemiological hypotheses. We illustrate our approach by focusing on the West Nile virus (WNV) spread in North America that has substantially impacted public, veterinary, and wildlife health. We apply an analytical workflow to a comprehensive WNV genome collection to test the impact of environmental factors on the dispersal of viral lineages and on viral population genetic diversity through time. We find that WNV lineages tend to disperse faster in areas with higher temperatures and we identify temporal variation in temperature as a main predictor of viral genetic diversity through time. By contrasting inference with simulation, we find no evidence for viral lineages to preferentially circulate within the same migratory bird flyway, suggesting a substantial role for non-migratory birds or mosquito dispersal along the longitudinal gradient.

Genomic surveillance reveals multiple introductions of SARS-CoV-2 into Northern California.

The coronavirus disease 2019 (COVID-19) pandemic caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has spread globally, with >365,000 cases in California as of 17 July 2020. We investigated the genomic epidemiology of SARS-CoV-2 in Northern California from late January to mid-March 2020, using samples from 36 patients spanning nine counties and the Grand Princess cruise ship. Phylogenetic analyses revealed the cryptic introduction of at least seven different SARS-CoV-2 lineages into California, including epidemic WA1 strains associated with Washington state, with lack of a predominant lineage and limited transmission among communities. Lineages associated with outbreak clusters in two counties were defined by a single base substitution in the viral genome. These findings support contact tracing, social distancing, and travel restrictions to contain the spread of SARS-CoV-2 in California and other states.