Evaluation of HIV- 1 Drug resistance Next Generation Sequencing Data Analysis Pipelines

Abstract

Next-generation sequencing (NGS) has transformed HIV drug resistance (HIVDR) testing by enhancing sensitivity for detecting drug resistance mutations (DRM). However, variability in the analytical outcomes generated by different bioinformatics pipelines for HIVDR interpretation raises concerns about the reliability and reproducibility of reported results. This study presents a secondary analysis of sequencing data from a 2019 South African HIVDR surveillance study, evaluating six widely used NGS analysis pipelines, PASeq, HyDRA, MiCall, DeepChek HIV, Stanford HIVdb-NGS (beta), and Exatype. Focusing on eight key mutations (K65R, K70R, K103N, V106M, M184V, G190A, K219E, and P225H), pipeline performance was compared under default and customized ‘Winnipeg’ parameter settings, assessing sensitivity, specificity, accuracy, and linearity. Across pipelines, sensitivity ranged from 92.3% to 100.0% and specificity from 96.4% to 100.0%, depending on the applied detection thresholds, while linearity analysis (R² values) demonstrated strong agreement (≥0.98) across all platforms. Results demonstrated notable variability in DRM detection, especially for low-frequency variants. Implementation of Winnipeg settings marginally improved concordance across pipelines, reducing discrepancies in DRM identification. This study highlights the significant impact of bioinformatics parameters on HIVDR reporting and advocates for standardized quality control measures to enhance consistency in NGS-based resistance testing. By identifying pipeline-specific biases and proposing quality control recommendations, the findings contribute to improving the reliability of genotypic HIVDR testing, with important implications for clinical management, surveillance programs, and optimization of HIV treatment strategies in resource-limited settings.