ABSTRACT
Sequencing of SARS-CoV-2 in wastewater provides a key opportunity to monitor the prevalence of variants spatiotemporally, potentially facilitating their detection simultaneously with, or even prior to, observation through clinical testing. However, there are multiple sequencing methodologies available. This study aimed to evaluate the performance of alternative protocols for detecting SARS-CoV-2 variants. We tested the detection of two synthetic RNA SARS-CoV-2 genomes in a wide range of ratios and at two concentrations representative of those found in wastewater using whole-genome and Spike-gene-only protocols utilising Illumina and Oxford Nanopore platforms. We developed a Bayesian hierarchical model to determine the predicted frequencies of variants and the error surrounding our predictions. We found that most of the sequencing protocols detected polymorphic nucleotide frequencies at a level that would allow accurate determination of the variants present at higher concentrations. Most methodologies, including the Spike-only approach, could also predict variant frequencies with a degree of accuracy in low-concentration samples but, as expected, with higher error around the estimates. All methods were additionally confirmed to detect the same prevalent variants in a set of wastewater samples. Our results provide the first quantitative statistical comparison of a range of alternative methods that can be used successfully in the surveillance of SARS-CoV-2 variant frequencies from wastewater.
ABSTRACT
Wastewater-based epidemiology (WBE) has been extensively used during the COVID-19 pandemic to detect and monitor the spread of the SARS-CoV-2 virus and its variants. It has also proven to be an excellent tool to complement and support insights gained from reported clinical data. Globally, many groups have developed bioinformatics pipelines to analyse sequencing data from wastewater. Accurate calling of mutations from RNA extracted from wastewater samples is key in supporting clinical data to make informed decisions on the prevalence of variants, as well as in the use of WBE as a molecular surveillance tool. However, wastewater samples can be challenging to extract and sequence, and performance of variant-calling algorithms in this context has, so far, not been investigated. Analysis of the data and assignment of circulating variants depends heavily on the accuracy of the variant caller, particularly given the degraded nature of the viral RNA and the heterogeneous nature of metagenomic samples. To address this, we compared the performance of six variant callers (VarScan, iVAR, GATK, FreeBayes, LoFreq and BCFtools), used widely in bioinformatics pipelines, on 19 synthetic samples with a known mix of three different SARS-CoV-2 variant genomes (Alpha, Beta and Delta), as well as 13 wastewater samples collected in London between the 15 th and 18 th December 2021. Using the Quasimodo benchmarking tool to compare the six variant callers, we assessed the fundamental parameters of recall (sensitivity) and precision (specificity) in confirming the presence of a variant within the population. Our results show that BCFtools, FreeBayes and VarScan called the expected mutations with higher precision and recall than iVAR or GATK, although the latter identified more expected defining mutations. LoFreq gave the least reliable results due to the high number of false positive mutations detected, resulting in lower precision. Similar results were obtained for both the synthetic and wastewater samples.