Background: Resistance to antiretroviral drugs is a complex and evolving area with relevant implications in the treatment of human immunodeficiency virus (HIV) infection. Several rules, algorithms and full-fledged computer programs have been developed to assist the HIV specialist in the choice of the best patient-tailored therapy.
Methods: Experts’ rules and statistical/machine learning algorithms for interpreting HIV drug resistance, along with their program implementations, were retrieved from PubMed and other on-line resources to be critically reviewed in terms of technical approach, performance, usability, update, and evolution (i.e. inclusion of novel drugs or expansion to other viral agents).
Results: Several drug resistance prediction algorithms for the nucleotide/nucleoside/non-nucleoside reverse transcriptase, protease and integrase inhibitors as well as coreceptor antagonists are currently available, routinely used, and have been validated thoroughly in independent studies. Computer tools that combine single-drug genotypic/phenotypic resistance interpretation and optimize combination antiretroviral therapy have been also developed and implemented as web applications. Most of the systems have been updated timely to incorporate new drugs and few have recently been expanded to meet a similar need in the Hepatitis C area. Prototype systems aiming at predicting virological response from both virus and patient indicators have been recently developed but they are not yet being routinely used.
Conclusion: Computing HIV genotype to predict drug susceptibility in vitro or response to combination antiretroviral therapy in vivo is a continuous and productive research field, translating into successful treatment decision support tools, an essential component of the management of HIV patients.
Keywords: Anti-retroviral therapy, drug resistance, expert system, HIV, statistical learning.