VIRALGUIDE: DRUG-CLASS-LEVEL HIV RESISTANCE PREDICTION USING BIDIRECTIONAL LSTM, XGBOOST ENSEMBLE AND CROSS-MODAL ATTENTION FUSION
DOI: 
SHARE THIS ARTICLE
DOWNLOAD ARTICLE
REVIEW REPORT
CITE THIS ARTICLE
Antiretroviral drug resistance is a persistent problem in achieving sustained viral suppression in diverse populations, and is a driver of HIV treatment failure. Current predictive models tend to categorize patients as responders or non-responders, an outcome that is too vague for clinicians to make meaningful decisions about treatment. In this paper we introduce ViralGuide, a tri-modal ensemble framework that independently estimates the probability of resistance to each of the four drug classes of interest (NRTIs, NNRTIs, Protease Inhibitors, INSTIs) via a combined processing of HIV genomic sequence in a Bidirectional LSTM network and structured Drug Resistance Mutation flags with normalized clinical parameters in gradient-boosted classifiers. Monte Carlo Dropout inference is used to estimate uncertainty and ensemble outputs are fed into a WHO guideline-aligned recommendation engine which outputs specific ART regimens with traceable clinical reasoning. Interpretability is tackled at the patient level through waterfall plots of SHAP values and counterfactual explanations of modifiable clinical features and longitudinal trajectory modelling over treatment visits. The findings reflect that the system can successfully improve discrimination rates per class as compared to the single-modality baseline systems, which make it a viable tool for linking resistance prediction to individual treatment planning for HIV treatment. In addition, the system enriches the predictions with clinically meaningful context features such as the level of HIV disease (based on CD4 levels) and an estimate of HIV subtype (based on the mutation pattern). These additions add to the interpretability and provide greater clinical context for decision-making.
Keywords
Article Information
Article Metrics
HOW TO CITE
Copy
Kausar, E. N. & Kandasamy, J. (2026). Viralguide: Drug-Class-Level HIV Resistance Prediction using Bidirectional LSTM, Xgboost Ensemble and Cross-Modal Attention Fusion. International Journal of Science, Strategic Management and Technology, 02(05). https://doi.org/10.55041/ijsmt.v2i5.237
Kausar, Eram, and Janaki Kandasamy. "Viralguide: Drug-Class-Level HIV Resistance Prediction using Bidirectional LSTM, Xgboost Ensemble and Cross-Modal Attention Fusion." International Journal of Science, Strategic Management and Technology, vol. 02, no. 05, 2026, pp. . doi:https://doi.org/10.55041/ijsmt.v2i5.237.
Kausar, Eram, and Janaki Kandasamy. "Viralguide: Drug-Class-Level HIV Resistance Prediction using Bidirectional LSTM, Xgboost Ensemble and Cross-Modal Attention Fusion." International Journal of Science, Strategic Management and Technology 02, no. 05 (2026). https://doi.org/https://doi.org/10.55041/ijsmt.v2i5.237.
References
[1] G. A. McComsey, M. Lingohr-Smith, R. Rogers, J. Lin, and P. Donga, "Adherence patterns to antiretroviral therapy among HIV-1 infected individuals across United States clinical settings," Advances in Therapy, vol. 38, no. 9, pp. 4961–4974, 2021.
[2] D. N. Mamo, T. M. Yilma, M. Fekadie, Y. Sebastian, T. Bizuayehu, M. S. Melaku, and A. D. Walle, "Applying machine learning classifiers to predict virological failure in HIV patients receiving antiretroviral treatment at a comprehensive hospital in Amhara Region, Ethiopia," BMC Medical Informatics and Decision Making, vol. 23, no. 1, p. 75, 2023.
[3] B. Bayu, A. Tariku, A. B. Bulti, Y. A. Habitu, T. Derso, and D. F. Teshome, "Clinical and behavioral determinants of virological failure in patients on highly active antiretroviral therapy: a case-control analysis," HIV/AIDS — Research and Palliative Care, pp. 153–159, 2017.
[4] G. Di Teodoro, M. Pirkl, F. Incardona, I. Vicenti, A. Sonnerborg, R. Kaiser, L. Palagi, M. Zazzi, and T. Lengauer, "Leveraging temporal mutation dynamics to improve antiretroviral therapy outcome prediction in HIV-1 infection," Bioinformatics, vol. 40, no. 6, 2024.
[5] B. T. Seboka, D. E. Yehualashet, and G. A. Tesfa, "Machine learning approaches for predicting viral load and CD4 cell count in people living with HIV on antiretroviral treatment in Gedeo Zone public hospitals," International Journal of General Medicine, pp. 435–451, 2023.
[6] M. Maskew, K. Sharpey-Schafer, L. De Voux, T. Crompton, J. Bor, M. Rennick, A. Chirowodza et al., "Predictive modeling for retention and viral load suppression in South African HIV treatment programs," Scientific Reports, vol. 12, no. 1, p. 12715, 2022.
[7] M. E. Ekpenyong, P. I. Etebong, and T. C. Jackson, "A fuzzy-multidimensional deep learning framework for predicting antiretroviral therapy response," Heliyon, vol. 5, no. 7, 2019.
[8] K. R. Bisaso, S. A. Karungi, A. Kiragga, J. K. Mukonzo, and B. Castelnuovo, "Comparative analysis of logistic regression-based machine learning models for predicting early virological suppression in antiretroviral-initiating HIV patients," BMC Medical Informatics and Decision Making, vol. 18, pp. 1–10, 2018.
[9] D. Bukenya, B. N. Mayanja, S. Nakamanya, R. Muhumuza, and J. Seeley, "Barriers to long-term antiretroviral therapy adherence among individuals with poor viral suppression in Uganda," AIDS Research and Therapy, vol. 16, pp. 1–9, 2019.
[10] Stanford University HIV Drug Resistance Database.
Ethics and Compliance
Indexed In
Similar Articles
Previous Article
