A2 Optimization of the results generated by large-scale sequencing for the study of drug resistance in HIV infection: A systematic review

Next-generation sequencing (NGS) approaches are now used in many clinical diagnostic laboratories for the routine diagnosis of resistance to antiretrovirals approved for the treatment of HIV infection. As some of NGS platforms may be a source of sequencing error, it is necessary to improve currently available protocols and implement bioinformatics tools that may help to correctly identify the presence of resistance mutations with clinical impact. In this study, we reviewed all studies dealing with software or methods aiming to decrease these errors, published during the period 2006–2016. We considered, as bioinformatic strategies, software aiming to delete or detect sequencing errors, and as protocol improvements, those changes in PCR temperature profiles and/or reagent concentration aiming to minimize sequencing errors. We used a combination of non-MeSH and MeSH terms related to error correction and NGS sequence filtering. All abstracts of papers available through January 2006 and June 2016 were reviewed. Our search identified 611 studies, we finally selected seven papers that met all the eligibility criteria, three of which dealt with protocol modifications and four with bioinformatics aiming to eliminate errors. Some studies are mainly focused on improving protocols for decreasing the magnitude of errors during the polymerase change reaction. Other studies propose specific bioinformatics tools, able to reach both a 93–98 per cent reduction of indels (insertions/deletions) and a sensitivity and specificity close to 100 per cent in single nucleotide polymorphism variant calling. Moreover, error rates decreased from a median value (95% CI) of 0.2 per cent (0.008–0.4) before processing to 0.06 per cent (0.05–0.08) after using a bioinformatic tool. All the software did not incur in a high loss in the number of reads. New protocols and bioinformatics tools that improve the accuracy of NGS results must be considered for correct analysis of HIV resistance mutations. We recommend using bioinformatic software to filter short and low-quality sequences, and using high fidelity polymerases.