DisCVR: Rapid viral diagnosis from high-throughput sequencing data

Abstract

12High-throughput sequencing (HTS) enables most pathogens in a clinical sample to be detected 13 from a single analysis, thereby providing novel opportunities for diagnosis, surveillance and epi-14 demiology. However, this powerful technology is difficult to apply in diagnostic laboratories be-15 cause of its computational and bioinformatic demands. We have developed DisCVR, which de-16 tects known human viruses in clinical samples by matching sample k-mers (22 nucleotide se-17 quences) to k-mers from taxonomically labelled viral genomes. DisCVR was validated using pub-18 lished HTS data for 89 clinical samples from adults with upper respiratory tract infections. These 19 samples had been tested for viruses metagenomically and also by real-time polymerase chain 20 reaction assay, which is the standard diagnostic method. DisCVR detected human viruses with 21 high sensitivity (79%) and specificity (100%), and was able to detect mixed infections. Moreover, 22 it produced results comparable to those in a published metagenomic analysis of 177 blood sam-23 ples from patients in Nigeria. DisCVR has been designed as a user-friendly tool for detecting hu-24 man viruses from HTS data using computers with limited RAM and processing power, and in-25 cludes a graphical user interface to help users interpret and validate the output. It is written in 26 Java and is publicly available from Issue Section: Resources 30 2 1. Introduction 31 The standard method for rapidly detecting known human viruses in clinical samples is the 32 polymerase chain reaction (PCR), in which short oligonucleotides are used to amplify and probe 33 specific regions of viral genomes. The limitations of this technique include the targeting of a rela-34 tively small number of viruses per assay and a dependence on sequence conservation among 35 viral strains. High-throughput sequencing (HTS) provides approaches to viral diagnosis that have 36 much greater scope. Thus, metagenomic analysis of HTS data can provide extensive viral geno-37 typing information, as well as the characterization of complex multiple infections (Thorburn et al. 38 2015). Several metagenomic pipelines using de novo assembly and homology matching have 39 been developed for virus detection (Li et al. 2016; Maarala et al. 2018; Ren et al. 2017; Scheuch 40 et al. 2015; Wang et al. 2013; Zheng et al. 2017). However, analysing HTS data using such ap-41 proaches brings heavy computing and bioinformatic demands that are difficult to meet and 42 standardize in diagnostic laboratories (Flygare et al. 2016). As a consequence, we have devel-43 oped DisCVR, which is a fast, accurate and easy-to-use tool for detecting known human viruses 44 in clinical samples. 45 DisCVR employs an abundance-based method, which is a metagenomic approach for 46 rapidly profiling the organisms present in a sample. It works by creating a database of short nu-47 cleotide sequences (k-mers) from a large set of viral reference sequences, tagging the k-mers 48 taxonomically according to the viruses from which they came, sc...