Shall I trust the report? Variable performance of Sanger sequencing revealed by deep sequencing on HIV drug resistance mutation detection

Background: The clinical utilisation of deep sequencing in HIV treatment has been hindered due to its unknown correlation with standard Sanger genotyping and the undetermined value of minority drug resistance mutation (DRM) detection. Objectives: To compare deep sequencing performance to standard Sanger genotyping with clinical samples, in an effort to delineate the correlation between the results from the two methods and to find the optimal deep sequencing threshold for clinical utilisation. Methods: We conducted a retrospective study using stored plasma collected from August 2014 to March 2018 for HIV genotyping with the commercial Sanger genotyping kit. Samples with available Sanger genotyping reports were further deep sequenced. Drug resistance was interpreted according to the Stanford HIV drug resistance database algorithm. Results: At 15–25% minority detection thresholds, 9–15% cases had underestimated DRMs by Sanger sequencing. The concordance between the Sanger and deep sequencing reports was 68–82% in protease-reverse transcriptase region and 88–97% in integrase region at 5–25% thresholds. The undetected drug resistant minority variants by Sanger sequencing contributed to the lower negative predictive value of Sanger genotyping in cases harbouring DRMs. Conclusions: Use of deep sequencing improved detection of antiretroviral resistance mutations especially in cases with virological failure or previous treatment interruption. Deep sequencing with 10–15% detection thresholds may be considered a suitable substitute for Sanger sequencing on antiretroviral DRM detection.