KG-Hub -- Building and Exchanging Biological Knowledge Graphs (preprint)

Knowledge graphs (KGs) are a powerful approach for integrating heterogeneous data and making inferences in biology and many other domains, but a coherent solution for constructing, exchanging, and facilitating the downstream use of knowledge graphs is lacking. Here we present KG-Hub, a platform that enables standardized construction, exchange, and reuse of knowledge graphs. Features include a simple, modular extract-transform-load (ETL) pattern for producing graphs compliant with Biolink Model (a high-level data model for standardizing biological data), easy integration of any OBO (Open Biological and Biomedical Ontologies) ontology, cached downloads of upstream data sources, versioned and automatically updated builds with stable URLs, web-browsable storage of KG artifacts on cloud infrastructure, and easy reuse of transformed subgraphs across projects. Current KG-Hub projects span use cases including COVID-19 research, drug repurposing, microbial-environmental interactions, and rare disease research. KG-Hub is equipped with tooling to easily analyze and manipulate knowledge graphs. KG-Hub is also tightly integrated with graph machine learning (ML) tools which allow automated graph machine learning, including node embeddings and training of models for link prediction and node classification.

The impact of environmental mycobiomes on geographic variation in COVID-19 mortality (preprint)

Mortality rates during the COVID-19 pandemic have varied by orders of magnitude across communities in the United States. Individual, socioeconomic, and environmental factors have been linked to health outcomes of COVID-19. It is now widely appreciated that the environmental microbiome, composed of microbial communities associated with soil, water, atmosphere, and the built environment, impacts immune system development and susceptibility to immune-mediated disease. The human microbiome has been linked to individual COVID-19 disease outcomes, but there are limited data on the influence of the environmental microbiome on geographic variation in COVID-19 across populations. To fill this knowledge gap, we used taxonomic profiles of fungal communities associated with 1,135 homes in 494 counties from across the United States in a machine learning analysis to predict COVID-19 Infection Fatality Ratios (the number of deaths caused by COVID-19 per 1000 SARS-CoV-2 infections; 'IFR'). Here we show that exposure to increased fungal diversity, and in particular indoor exposure to outdoor fungi, is associated with reduced SARS-CoV-2 IFR. Further, we identify seven fungal genera that are the predominant drivers of this protective signal and may play a role in suppressing COVID-19 mortality. This relationship is strongest in counties where human populations have remained stable over at least the previous decade, consistent with the importance of early-life microbial exposures. We also assessed the explanatory power of 754 other environmental and socioeconomic factors, and found that indoor-outdoor fungal beta-diversity is amongst the strongest predictors of county-level IFR, on par with the most important known COVID-19 risk factors, including age. We anticipate that our study will be a starting point for further integration of environmental mycobiome data with population health information, providing an important missing link in our capacity to identify vulnerable populations. Ultimately, our identification of specific genera predicted to be protective against COVID-19 mortality may point toward novel, proactive therapeutic approaches to infectious disease.

Estimating area-level variation in SARS-CoV-2 infection fatality ratios (preprint)

Background: During a pandemic, estimates of geographic variability in disease burden are important but limited by the availability and quality of data. Methods: We propose a framework for estimating geographic variability in testing effort, total number of infections, and infection fatality ratio (IFR). Because symptomatic people are more likely to seek testing, we use a noncentral hypergeometric model that accounts for differential probability of positive tests. We apply this framework to the United States (U.S.) COVID-19 pandemic to estimate county-level SARS-CoV-2 IFRs from March 1, 2020 to October 31, 2020. Using data on population size, number of observed cases, number of reported deaths in each U.S. county and state, and number of tests in each U.S. state, we develop a series of estimators to identify the number of SARS-CoV-2 infections and IFRs at the county level. We then perform a simulation and compare the estimated values to simulated values to demonstrate the validity of our approach. Findings: Applying the county-level estimators to the real, unsimulated COVID-19 data spanning March 1, 2020 to October 31, 2020 from across the U.S., we found that IFRs varied from 0 to 0.0273, with an interquartile range of 0.0022 and a median of 0.0018. The estimators for IFRs, number of infections, and number of tests showed high accuracy and precision; for instance, when applied to simulated validation data sets, across counties, Pearson correlation coefficients between estimator means and true values were 0.88, 0.95, and 0.74, respectively. Interpretation: We propose an estimation framework that can be used to identify area-level variation in IFRs and performs well to estimate county-level IFRs in the U.S. COVID-19 pandemic.

KG-COVID-19: a framework to produce customized knowledge graphs for COVID-19 response (preprint)

Integrated, up-to-date data about SARS-CoV-2 and coronavirus disease 2019 (COVID-19) is crucial for the ongoing response to the COVID-19 pandemic by the biomedical research community. While rich biological knowledge exists for SARS-CoV-2 and related viruses (SARS-CoV, MERS-CoV), integrating this knowledge is difficult and time consuming, since much of it is in siloed databases or in textual format. Furthermore, the data required by the research community varies drastically for different tasks - the optimal data for a machine learning task, for example, is much different from the data used to populate a browsable user interface for clinicians. To address these challenges, we created KG-COVID-19, a flexible framework that ingests and integrates biomedical data to produce knowledge graphs (KGs) for COVID-19 response. This KG framework can also be applied to other problems in which siloed biomedical data must be quickly integrated for different research applications, including future pandemics. BIGGER PICTUREAn effective response to the COVID-19 pandemic relies on integration of many different types of data available about SARS-CoV-2 and related viruses. KG-COVID-19 is a framework for producing knowledge graphs that can be customized for downstream applications including machine learning tasks, hypothesis-based querying, and browsable user interface to enable researchers to explore COVID-19 data and discover relationships.

A Lymph Node Targeted Amphiphile Vaccine Induces Potent Cellular and Humoral Immunity to SARS-CoV-2 (preprint)

The SARS-CoV-2 pandemic has led to public health, economic, and social consequences that mandate urgent development of effective vaccines to contain or eradicate infection. To that end, we evaluated a novel amphiphile (AMP) vaccine adjuvant, AMP-CpG, composed of diacyl lipid-modified CpG, admixed with the SARS-CoV-2 Spike-2 receptor binding domain protein as a candidate vaccine (ELI-005) in mice. AMP immunogens are efficiently delivered to lymph nodes, where innate and adaptive immune responses are generated. Compared to alum, AMP immunization induced >25-fold higher antigen-specific T cells which produced multiple Th1 cytokines and trafficked into lung parenchyma and respiratory secretions. Antibody responses favored Th1 isotypes (IgG2bc, IgG3) and potently neutralized Spike-2-ACE2 receptor binding, with titers 265-fold higher than the natural immune response from convalescent COVID-19 patients; responses were maintained despite 10-fold dose-reduction in Spike antigen. Both cellular and humoral immune responses were preserved in aged mice. These advantages merit clinical translation to SARS-CoV-2 and other protein subunit vaccines.