Search

Leveraging prior viral genome sequencing data to make predictions on whether an unknown, emergent virus harbors a ‘phenotype-of-concern’ has been a long-sought goal of genomic epidemiology. A predictive phenotype model built from nucleotide-level information alone has previously been considered un-tenable with respect to RNA viruses due to the ultra-high intra-sequence variance of their genomes, even within closely related clades. Building from our prior work developing a degenerate k-mer method to accommodate this high intra-sequence variation of RNA virus genomes for modeling frameworks, and leveraging a taxonomic ‘group-shuffle-split’ paradigm on complete coronavirus assemblies from prior to October 2018, we trained multiple regularized logistic regression classifiers at the nucleotide k-mer level capable of accurately predicting withheld SARS-CoV-2 genome sequences as human pathogens and accurately predicting withheld Swine Acute Diarrhea Syndrome coronavirus (SADS-CoV) genome sequences as non-human pathogens. LASSO feature selection identified several degenerate nucleotide predictor motifs with high model coefficients for the human pathogen class that were present across widely disparate classes of coronaviruses. However, these motifs differed in which genes they were present in, what specific codons were used to encode them, and what the translated amino acid motif was. This emphasizes the importance of a phenetic view of emerging pathogenic RNA viruses, as opposed to the canonical phylogenetic interpretations most-commonly used to track and manage viral zoonoses. Applying our model to more recent Orthocoronavirinae genomes deposited since October 2018 yields a novel contextual view of pathogen-potential across bat-related, canine-related, porcine-related, and rodent-related coronaviruses and critical adaptations which may have contributed to the emergence of the pandemic SARS-CoV-2 virus. Finally, we discuss the utility of these predictive models (and their associated predictor motifs) to novel biosurveillance protocols that substantially increase the ‘pound-for-pound’ information content of field-collected sequencing data and make a strong argument for the necessity of routine collection and sequencing of zoonotic viruses.

Metagenomics is emerging as an important tool in biosurveillance, public health, and clinical applications. However, ease-of-use for execution and data analysis remains a barrier-of-entry to the adoption of metagenomics in applied health and forensics settings. In addition, these venues often have more stringent requirements for reporting, accuracy, and precision than the traditional ecological research role of the technology. Here, we present PanGIA (Pan-Genomics for Infectious Agents), a novel bioinformatics analysis platform for hosting, processing, analyzing, and reporting shotgun metagenomics data of complex samples suspected of containing one or more pathogens. PanGIA was developed to address gaps that often preclude clinicians, medical technicians, forensics personnel, or other non-expert end-users from the routine application of metagenomics for pathogen identification. Though primarily designed to detect pathogenic microorganisms within clinical and environmental metagenomics data, PanGIA also serves as an analytical framework for microbial community profiling and comparative metagenomics. To provide statistical confidence in PanGIA’s taxonomic assignments, the system provides two independent estimations of probability for species and strain level detection. First, PanGIA integrates coverage data with ‘uniqueness’ information mapped across each reference genome for a stand-alone determination of confidence for each query sequence at each taxonomy level. Second, if a negative-control sample is provided, PanGIA compares this sample with a corresponding experimental unknown sample and determines a measure of confidence associated with ‘detection above background’. An integrated graphical user interface allows interactive interrogation and enables users to summarize multiple sample results by confidence score, normalized read abundance, reference genome linear coverage, depth-of-coverage, RPKM, and other metrics to detect specific organisms-of-interest. Comparison testing of the PanGIA algorithm against a number of recent k-mer, read-mapping, and marker-gene based taxonomy classifiers across various real-world datasets with spiked targets shows superior mean positive predictive value, sensitivity, and specificity. PanGIA can process a five million paired-end read dataset in under 1 hour on commodity computational hardware. The source code and documentation are publicly available at https://github.com/LANL-Bioinformatics/PanGIA or https://github.com/mriglobal/PanGIA. The database for PanGIA can be downloaded from ftp://bioinformatics.mriglobal.org/. The full GUI-based PanGIA analysis environment is available in a Docker container and can be installed from https://hub.docker.com/r/poeli/pangia/.

SUMMARYIn the analysis of genomic sequence data, so-called “alignment free” approaches are often selected for their relative speed compared to alignment-based approaches, especially in the application of distance comparisons and taxonomic classification1,2,3,4. These methods are typically reliant on excising K-length substrings of the input sequence, called K-mers5. In the context of machine learning, K-mer based feature vectors have been used in applications ranging from amplicon sequencing classification to predictive modeling for antimicrobial resistance genes6,7,8. This can be seen as an analogy of the “bag-of-words” model successfully employed in natural language processing and computer vision for document and image classification9,10. Feature extraction techniques from natural language processing have previously been analogized to genomics data11; however, the “bag-of-words” approach is brittle in the RNA virus space due to the high intersequence variance and the exact matching requirement of K-mers. To reconcile the simplicity of “bag-of-words” methods with the complications presented by the intrinsic variance of RNA virus space, a method to resolve the fragility of extracted K-mers in a way that faithfully reflects an underlying biological phenomenon was devised. Our algorithm, Vorpal, allows the construction of interpretable linear models with clustered, representative ‘degenerate’ K-mers as the input vector and, through regularization, sparse predictors of binary phenotypes as the output. Here, we demonstrate the utility of Vorpal by identifying nucleotide-level genomic motif predictors for binary phenotypes in three separate RNA virus clades; human pathogen vs. non-human pathogen in Orthocoronavirinae, hemorrhagic fever causing vs. non-hemorrhagic fever causing in Ebolavirus, and human-host vs. non-human host in Influenza A. The capacity of this approach for in silico identification of hypotheses which can be validated by direct experimentation, as well as identification of genomic targets for preemptive biosurveillance of emerging viruses, is discussed. The code is available for download at https://github.com/mriglobal/vorpal.

The future of infectious disease surveillance and outbreak response is trending towards smaller hand-held solutions for point-of-need pathogen detection.1–4Although recent advances have paved the way for these technologies to include sequencing of pathogens directly from clinical samples, the ability to carry out unbiased sequencing for pathogen discovery and subtyping directly from environmental samples has yet to be demonstrated with hand-held platforms.5Products such as the two3 qPCR system from Biomeme Inc., as well as the MinION from Oxford Nanopore Technologies, have generated renewed prospects for point-of-need diagnostics and near real-time environmental testing and characterization of viral and microbial pathogens. Here, samples ofCulex cedeceimosquitoes collected in Southern Florida, USA were tested for Venezuelan Equine Encephalitis Virus (VEEV), a previously-weaponized arthropod-borne RNA-virus capable of causing acute and fatal encephalitis in animal and human hosts. A single 20-mosquito pool tested positive for VEEV by real-time reverse transcription quantitative PCR (RT-qPCR) on the Biomeme two3. The virus-positive sample was then subjected to unbiased metatranscriptome sequencing on the MinION and determined to contain Everglades Virus (EVEV), a strain of VEEV transmitted exclusively byCulex cedeceiin South Florida. The result was confirmed on “gold standard” thermocyclers and sequencing machines, and comparison to nanopore results is discussed. Our results demonstrate, for the first time, the use of unbiased sequence-based detection and subtyping of a high-consequence biothreat pathogen directly from an environmental sample using field-forward hardware and protocols. The further development and validation of methods designed for field-based diagnostic metagenomics and pathogen discovery, such as those suitable for use in mobile “pocket laboratories”, will address a growing demand for public health teams to carry out their mission where it is most urgent: at the point-of-need.6