Biology subjects

Roder, A.

Publications and source records attributed to Roder, A..

2 recordsLinked to original sources

bioRxiv2021

Accurate Virus Identification with Interpretable Raman Signatures by Machine Learning

Rapid identification of newly emerging or circulating viruses is an important first step toward managing the public health response to potential outbreaks. A portable virus capture device coupled with label-free Raman Spectroscopy holds the promise of fast detection by rapidly obtaining the Raman signature of a virus followed by a machine learning approach applied to recognize the virus based on its Raman spectrum. In this paper, we present a machine learning analysis on Raman spectra of human and avian viruses. A Convolutional Neural Network (CNN) classifier specifically designed for spectral data achieves very high accuracy for a variety of virus type or subtype identification tasks. In particular, it achieves 99% accuracy for classifying influenza virus type A vs. type B, 96% accuracy for classifying four subtypes of influenza A, 95% accuracy for differentiating enveloped and non-enveloped viruses, and 99% for differentiating avian coronavirus (infectious bronchitis virus, IBV) from other avian viruses. Furthermore, interpretation of neural net responses in the trained CNN model using a full-gradient algorithm highlights Raman spectral ranges that are most important to virus identification. By correlating ML-selected salient Raman ranges with the signature ranges of known biomolecules and chemical functional groups (e.g. amide, amino acid, carboxylic acid) we verify that our ML model effectively recognizes the Raman signatures of proteins, lipids and other vital functional groups present in different viruses and uses a weighted combination of these signatures to identify viruses. The accurate and interpretable machine learning model developed for Raman virus identification presents promising potential in a real-time virus detection system. Significance StatementA portable micro-fluidic platform for virus capture promises rapid enrichment and label-free optical identification of viruses by Raman spectroscopy. A large Raman dataset collected on a variety of viruses enables the training of machine learning (ML) models capable of highly accurate and sensitive virus identification. The trained ML models can then be integrated with the portable device to provide real-time virus detection and identification capability. We validate this conceptual framework by presenting highly accurate virus type and subtype identification results using a convolutional neural network to classify Raman spectra of viruses.

biophysics↗

bioRxiv2021

Diversity and selection of SARS-CoV-2 minority variants in the early New York City outbreak

High error rates of viral RNA-dependent RNA polymerases lead to diverse intra-host viral populations during infection. Errors made during replication that are not strongly deleterious to the virus can lead to the generation of minority variants. However, accurate detection of minority variants in viral sequence data is complicated by errors introduced during sample preparation and data analysis. We used synthetic RNA controls and simulated data to test seven variant calling tools across a range of allele frequencies and simulated coverages. We show that choice of variant caller, and use of replicate sequencing have the most significant impact on single nucleotide variant (SNV) discovery and demonstrate how both allele frequency and coverage thresholds impact both false discovery and false negative rates. We use these parameters to find minority variants in sequencing data from SARS-CoV-2 clinical specimens and provide guidance for studies of intrahost viral diversity using either single replicate data or data from technical replicates. Our study provides a framework for rigorous assessment of technical factors that impact SNV identification in viral samples and establishes heuristics that will inform and improve future studies of intrahost variation, viral diversity, and viral evolution. IMPORTANCEWhen viruses replicate inside a host, the virus replication machinery makes mistakes. Over time, these mistakes create mutations that result in a diverse population of viruses inside the host. Mutations that are neither lethal to the virus, nor strongly beneficial, can lead to minority variants that are minor members of the virus population. However, preparing samples for sequencing can also introduce errors that resemble minority variants, resulting in inclusion of false positive data if not filtered correctly. In this study, we aimed to determine the best methods for identification and quantification of these minority variants by testing the performance of seven commonly used variant calling tools. We used simulated and synthetic data to test their performance against a true set of variants, and then used these studies to inform variant identification in data from clinical SARS-CoV-2 clinical specimens. Together, analyses of our data provide extensive guidance for future studies of viral diversity and evolution.

microbiology↗