Inference of genetic relatedness between viral quasispecies from sequencing data

Abstract

BackgroundRNA viruses such as HCV and HIV mutate at extremely high rates, and as a result, they exist in infected hosts as populations of genetically related variants. Recent advances in sequencing technologies make possible to identify such populations at great depth. In particular, these technologies provide new opportunities for inference of relatedness between viral samples, identification of transmission clusters and sources of infection, which are crucial tasks for viral outbreaks investigations.ResultsWe present (i) an evolutionary simulation algorithm Viral Outbreak InferenCE (VOICE) inferring genetic relatedness, (ii) an algorithm MinDistB detecting possible transmission using minimal distances between intra-host viral populations and sizes of their relative borders, and (iii) a non-parametric recursive clustering algorithm Relatedness Depth (ReD) analyzing clusters’ structure to infer possible transmissions and their directions. All proposed algorithms were validated using real sequencing data from HCV outbreaks.ConclusionsAll algorithms are applicable to the analysis of outbreaks of highly heterogeneous RNA viruses. Our experimental validation shows that they can successfully identify genetic relatedness between viral populations, as well as infer transmission clusters and outbreak sources.

Analysis of Heterogeneous Genomic Samples Using Image Normalization and Machine Learning

“…However, this measure is sensitive to noise and presence of outliers, and does not take into account the whole population structure. Recently, several simulation-based and network-based distance measures have been proposed [13,32], which overcome above-mentioned limitations at the cost of lesser scalability. Thus, the universal, accurate and efficiently computable inter-population distance measure, which takes into account the complex population structures still has to be developed.…”

“…Until recently, most available tools for outbreak investigations analyzed only a single representative sequence per population (usually consensus sequence) [34,35]. Although several recently published tools allow to take into account entire intra-host populations [13,32,36], the problem of comparison and clustering of viral populations still remains challenging [15].…”

“…The data [4] used for classification of intrahost HCV populations as recent and chronic consists of 365 NGS samples, including 108 datasets correponding to recently infected hosts and 257 datasets belonging to chronically infected hosts. For clustering and indentification of outbreaks we used the benchmark dataset [8,13,32] that consists of HCV intra-host populations collected from 335 infected individuals by Centers for Disease Control and Prevention in 2008-2013. Of these, 142 HCV samples belong to epidemiologically curated outbreaks involving from 2 to 19 patients, while the remaining datasets are epidemiologically isolated cases.…”

“…All these measures range from 0 to 1 and the values closer to 1 refer to better clustering efficiency. To evaluate the effectiveness of the normalization method in detecting relatedness between any pair of samples, we compute AUROC (Area under ROC curve) is computed (as done in [13]). Viral populations taken from the same outbreak are considered as genetically related, otherwise as unrelated.…”

“…Modern computational virology continues mostly to rely on classical approaches, which includes sequence analysis, phylogenetics/phylodynamics and structural bioinformatics [21,22]. In the recent years, these approaches started to be complemented with the network analysis [6,13,32]. Significant number of computational molecular epidemiology problems could be defined using classification-or clustering-based objective.…”