Machine learning identifies potential accessory resistance-associated mutations in HIV-1 integrase.
6 affiliations
- 1Department of General Virology, Uganda Virus Research Institute (UVRI), Entebbe, Uganda. [email protected].
- 2Department of General Virology, Uganda Virus Research Institute (UVRI), Entebbe, Uganda.
- 3Medical Research Council (MRC)/UVRI and London School of Hygiene and Tropical Medicine (LSHTM) Uganda Research Unit, Entebbe, Uganda.
- 4Department of Immunology and Molecular Biology, College of Health Sciences, Makerere University, Kampala, Uganda.
- 5African Center of Excellence in Bioinformatics and Data Intensive Sciences, Infectious Diseases Institute, Makerere University, Kampala, Uganda.
- 6Department of Computer Science, College of Computing and Information Sciences, Makerere University, Kampala, Uganda.
Abstract
Background
Although integrase strand transfer inhibitors (INSTIs) have a high genetic barrier to resistance, cases of virological failure continue to emerge, sometimes in the absence of major resistance-associated mutations. Conventional genotypic and phenotypic resistance testing is costly and time-intensive, especially in resource-limited settings, and is inherently limited in detecting novel resistance pathways in any setting. Machine learning offers a scalable approach to uncover previously unrecognized resistance-associated patterns in HIV-1 genomic data.Results
We analyzed 41,247 publicly available HIV-1 integrase sequences from ART-naïve and ART-experienced individuals using interpretable machine learning algorithms. Extreme Gradient Boosting (XGBoost), Gradient Boosting Machines (GBM), Random Forests (RF), Support Vector Machines (SVM), Logistic Regression (LR), and Naïve Bayes classifiers were trained to distinguish treatment status based solely on HIV-1 integrase mutation profiles. XGBoost outperformed other classifiers, with an AUROC of 0.92. Top-ranking mutations identified by the classifiers, including S283G, T112V, D278A, K136Q, T125A, V201I, V31I, I72V, A265V, G134N, and I135V among others, were significantly more prevalent in ART-experienced sequences. Structural analysis showed that most mutations potentially destabilize the three-dimensional structure of HIV-1 integrase. Relative risk (RR) analysis identified twenty three significant co-occurring mutation pairs with major INSTI resistance mutations, including G118R-R269K (RR = 2.3), Q148H-T125A (RR = 1.6), and Y143A-I135V (RR = 2.3). Most of these associations clustered within established resistance pathways (G118R, Q148/G140, Y143, and N155).Conclusions
Machine learning identified potential accessory resistance-associated mutations in HIV-1 integrase beyond established INSTI resistance pathways. These mutations may contribute to INSTI resistance via epistatic interactions with known major resistance mutations. Validation of the predictive framework using independent clinical datasets, together with experimental and longitudinal studies, is required to determine the functional impact of these mutations and their relevance to treatment outcomes.
Related
- Cancer type-specific profiling of chromatin regulator mutations identifies recurrent and co-occurring gene sets associated with improved survival following immune checkpoint therapy
- SMARCA4 MUTATIONS PROMOTE CHEMOTHERAPY RESISTANCE IN T-ACUTE LYMPHOBLASTIC LEUKEMIA THROUGH ALTERED CHROMATIN REMODELING AT AP1-ASSOCIATED REGULATORY REGIONS
- Pan-cancer analysis identifies ANGPT4 as a potential prognostic biomarker associated with the tumor microenvironment in kidney renal clear cell carcinoma
- Relative Contributions of Quinolone Resistance-Determining Region Mutations and Efflux Pump-Associated Mechanisms to Fluoroquinolone Resistance in Staphylococcus aureus Clinical Isolates and Their Associations with Virulence Gene Profiles.
- Molecular Surveillance of Antimalarial Drug Resistance-Associated Mutations in P. falciparum Imported from Africa into Guangdong Province, China, 2023-2025.
BibTeX
@article{ssekagiri2026machine,
title = {{Machine learning identifies potential accessory resistance-associated mutations in HIV-1 integrase.}},
author = {Alfred Ssekagiri and Deogratius Ssemwanga and David Patrick Kateete and Daudi Jjingo},
journal = {BMC genomics},
year = {2026},
month = aug,
doi = {10.1186/s12864-026-13243-3},
url = {https://europepmc.org/article/MED/42958478},
note = {CC BY},
}
RIS
TY - JOUR
TI - Machine learning identifies potential accessory resistance-associated mutations in HIV-1 integrase.
AU - Alfred Ssekagiri
AU - Deogratius Ssemwanga
AU - David Patrick Kateete
AU - Daudi Jjingo
PY - 2026
DA - 2026/08/03
JO - BMC genomics
DO - 10.1186/s12864-026-13243-3
UR - https://europepmc.org/article/MED/42958478
AB - Background Although integrase strand transfer inhibitors (INSTIs) have a high genetic barrier to resistance, cases of virological failure continue to emerge, sometimes in the absence of major resistance-associated mutations. Conventional genotypic and phenotypic resistance testing is costly and time-intensive, especially in resource-limited settings, and is inherently limited in detecting novel resistance pathways in any setting. Machine learning offers a scalable approach to uncover previously unrecognized resistance-associated patterns in HIV-1 genomic data.Results We analyzed 41,247 publicly available HIV-1 integrase sequences from ART-naïve and ART-experienced individuals using interpretable machine learning algorithms. Extreme Gradient Boosting (XGBoost), Gradient Boosting Machines (GBM), Random Forests (RF), Support Vector Machines (SVM), Logistic Regression (LR), and Naïve Bayes classifiers were trained to distinguish treatment status based solely on HIV-1 integrase mutation profiles. XGBoost outperformed other classifiers, with an AUROC of 0.92. Top-ranking mutations identified by the classifiers, including S283G, T112V, D278A, K136Q, T125A, V201I, V31I, I72V, A265V, G134N, and I135V among others, were significantly more prevalent in ART-experienced sequences. Structural analysis showed that most mutations potentially destabilize the three-dimensional structure of HIV-1 integrase. Relative risk (RR) analysis identified twenty three significant co-occurring mutation pairs with major INSTI resistance mutations, including G118R-R269K (RR = 2.3), Q148H-T125A (RR = 1.6), and Y143A-I135V (RR = 2.3). Most of these associations clustered within established resistance pathways (G118R, Q148/G140, Y143, and N155).Conclusions Machine learning identified potential accessory resistance-associated mutations in HIV-1 integrase beyond established INSTI resistance pathways. These mutations may contribute to INSTI resistance via epistatic interactions with known major resistance mutations. Validation of the predictive framework using independent clinical datasets, together with experimental and longitudinal studies, is required to determine the functional impact of these mutations and their relevance to treatment outcomes.
SN - 1471-2164
ER -