Sequence and structure based models of HIV-1 protease and reverse transcriptase drug resistance

Area of Science:

• Computational biology
• Virology
• Drug discovery

Background:

• Management of chronic human immunodeficiency virus type 1 (HIV-1) infection relies on antiretroviral medications.
• Drug-resistant mutations in HIV-1 protease (PR) and reverse transcriptase (RT) proteins can compromise treatment efficacy.
• Mutational patterns exhibit a spectrum of resistance, from susceptibility to cross-resistance across multiple drugs.

Purpose of the Study:

• To develop and implement statistical learning algorithms for predicting the impact of mutations in HIV-1 PR and RT on drug resistance.
• To create structure- and sequence-based models for forecasting resistance to multiple antiretroviral inhibitors.

Main Methods:

• Utilized a four-body statistical potential to represent mutant proteins as feature vectors based on environmental perturbations.
• Developed sequence-based models using n-grams (relative frequencies or counts) to generate attribute vectors for mutant proteins.
• Employed tenfold cross-validation to evaluate model performance using publicly available in vitro drug resistance data.

Main Results:

• Developed novel structure- and sequence-based predictive models for HIV-1 PR and RT drug resistance.
• Achieved competitive performance compared to existing sequence-based strategies.
• Demonstrated the ability of models to quantify the effects of mutations on inhibitor susceptibility.

Conclusions:

• The developed models offer orthogonal and complementary prediction methodologies for HIV-1 drug resistance.
• Introduced a novel technique for identifying effective or detrimental combinations of reverse transcriptase inhibitors for combination therapy.
• These models can aid in optimizing antiretroviral treatment strategies and predicting drug efficacy.