Current Approaches in Computational Drug Resistance Prediction in HIV

Background: Today a broad range of antiretroviral drug regimens are applicable for the successful suppression of virus replication in human immunodeficiency virus (HIV) infected people. However, there still remains an obstacle in therapy: the high mutation rate of the HI virus under drug pressure leads to resistant variants causing failure of permanent and effective treatment. Therefore, resistance testing is therefore inevitable to administer appropriate antiviral drugs to infected patients. Methods: By means of current high-throughput sequencing technologies, computational models have recently constituted important assistance in drug resistance prediction and can guide the choice of medical treatment. Several machine learning algorithms, e.g. support-vector machines, random forests, as well as statistical methods have been already applied to genotypic data and structural information to predict drug resistance. Results: In this review, we provide an overview of existing approaches in computational drug resistance prediction in HIV. We further highlight the challenges and limitations of current methods, e.g. time complexity and prediction of non-B subtypes. Conclusion: Moreover, we give a perspective on multi-label and multi-instance classification techniques that potentially tackle the problem of cross-resistances among drugs.