Quality control of low-frequency variants in SARS-CoV-2 genomes

Abstract

During the current outbreak of COVID-19, research labs around the globe submit sequences of the local SARS-CoV-2 genomes to the GISAID database to provide a comprehensive analysis of the variability and spread of the virus during the outbreak. We explored the variations in the submitted genomes and found a significant number of variants that can be seen only in one submission (singletons). While it is not completely clear whether these variants are erroneous or not, these variants show lower transition/transversion ratio. These singleton variants may influence the estimations of the viral mutation rate and tree topology. We suggest that genomes with multiple singletons even marked as high-covered should be considered with caution. We also provide a simple script for checking variant frequency against the database before submission.

Stability of SARS-CoV-2 Phylogenies

“…It can be difficult to distinguish sequencing errors of different types from genuine transmitted and non-transmitted mutations in genome sequences. Taking a conservative approach, many researchers remove mutations that are observed only once during the evolution of the virus when constructing a phylogenetic tree, as these may be more likely to be errors [21,22] , or non-transmitted mutations. However, systematic errors, where the same error from a common source is introduced many times in otherwise distinct viral genome sequences, are not removed by that approach [23,24] .…”

“…There are at least four possible sources of (real or apparent) mutations that recur within independent lineages in a tree, and each makes distinct predictions about the source of recurrent mutations (Table 1). In particular, recent work has shown a strong bias towards C>U mutation in the SARS-CoV-2 genome [21,[42][43][44] . Systematic errors, which usually result from consistent errors in molecular biology techniques or bioinformatic data data processing, need not reflect this bias and are not subject to natural selection.…”

Stability of SARS-CoV-2 phylogenies

“…It can be difficult to distinguish sequencing errors of different types from genuine transmitted and non-transmitted mutations in genome sequences. Taking a conservative approach, many researchers remove variants that are observed only once during the evolution of the virus when constructing a phylogenetic tree, as these may be more likely to be errors [22,23] or non-transmitted mutations. However, systematic errors, where the same error from a common source is introduced many times in otherwise distinct viral genome sequences, are not removed by that approach [24,25].…”

Peer Review #3 of "Excessive G–U transversions in novel allele variants in SARS-CoV-2 genomes (v0.2)"

“…P-values are based on a two-tailed Fishers exact test. Rayko et al have reported a lower transition/transversion ratio in singleton SARS-CoV-2 genomic variations (that can only be seen in one genome submission) [13]. As a separate control, we looked at mutations that were identified in two or more independent genome assemblies.…”