Fact and Fiction about 1%: Next Generation Sequencing and the Detection of Minor Drug Resistant Variants in HIV-1 Populations with and without Unique Molecular Identifiers

Abstract

Next generation sequencing (NGS) platforms have the ability to generate almost limitless numbers of sequence reads starting with a PCR product. This gives the illusion that it is possible to analyze minor variants in a viral population. However, including a PCR step obscures the sampling depth of the viral population, the key parameter needed to understand the utility of the data set for finding minor variants. Also, these high throughput sequencing platforms are error prone at the level where minor variants are of interest, confounding the interpretation of detected minor variants. A simple strategy has been applied in multiple applications of NGS to solve these problems. Prior to PCR, individual molecules are “tagged” with a unique molecular identifier (UMI) that can be used to establish the actual sample size of viral genomes sequenced after PCR and sequencing. In addition, since PCR generates many copies of each sequence tagged to a specific UMI, a template consensus sequence (TCS) can be created from the many reads of each template, removing virtually all of the method error. From this perspective we examine our own use of a UMI, called Primer ID, in the detection of minor drug resistant variants in HIV-1 populations.

HIV-1 drug resistance and genetic diversity in a cohort of people with HIV-1 in Nigeria

“…at 100%) is defined by the viral sequence population's sampling depth. Grouping specimens based on the sequence sampling depth, and thus the levels of sensitivity of minor variants, is therefore, essential [ 49 ]. The PID sequencing allows us to tag each sequenced viral RNA with a unique molecular ID (UMI) and the TCS number of one specimen represents the sampling depths.…”

An assessment of Nano-RECall: Interpretation of Oxford Nanopore sequence data for HIV-1 drug resistance testing

“…The strength of this comparison was that we compared sequencing results using the same PCR amplicon. This eliminated the influence of PCR resampling error and PCR introduced mutations on the findings [13]. It therefore enabled us to assess the impact of the sequencing platforms, automated Sanger sequencing versus Oxford Nanopore Flongle sequencing coupled with their respective bioinformatics pipelines.…”

“…However, the clinical benefit of this over and above detection of majority variants may be limited to particular settings such as pre-treatment drug resistance assessment or when using etravirine in treatment-experienced patients [17][18][19]. Moreover, PCR resampling error and sequencing error compound at low frequency variant thresholds so that the lower the threshold the higher the uncertainty of measurement and the more unclear whether these mutations are of clinical significance [13]. Approaches to improve the accuracy such as using random primer IDs, are cumbersome from a workflow and bioinformatics approach [20,21].…”

Application of a Sanger-Based External Quality Assurance Strategy for the Transition of HIV-1 Drug Resistance Assays to Next Generation Sequencing

“…While the Sanger-based detection of HIV DRMs was based on a pre-established cutoff defined by the early manufacturers of those kits, no pre-defined detection thresholds are defined for NGS. Zhou and Swanstrom suggest that using a 1% threshold for NGS is unrealistic due to the inherent frequency of errors generated during PCR, especially in assays that do not control for sampling depth within their assays [ 32 ]. Becker et al claim that drug resistance can be reported with an accuracy from 2–100% based on testing done in two laboratories [ 33 ].…”