Analysis of Resistance to Human Immunodeficiency Virus Protease Inhibitors Using Molecular Mechanics and Machine Learning Strategies

Abstract

Problem statement: Drug resistance is the most important factor, which influences the failure of current HIV therapies. So, the ability to predict the drug resistance of HIV-1 protease mutants will be useful in developing more effective and longer lasting treatment regimens. Approach: Drug resistance of HIV-1 protease is predicted with two current protease inhibitors (Indinavir and Saquinavir). The problem was approached from two perspectives. First, structural features of the HIV protease with inhibitor complex were constructed. Next, a classifier was constructed based on the patterns of various drug resistant mutants. In first stage SPDB viewer (for making mutations) and INSIGHT II (for analyzing binding energies and hydrogen bond contact with the inhibitor and the binding site) software's were used for structural property analysis. In the second stage a supervised learning linear Classifier (SVM-LIB) in DTREG tool has been used to analyze the Resistant and susceptible patterns. Finally Genetic Algorithm in Matlab tool has been used for Optimization. Results: Structural data mining performed linear SVM model gives 93% accuracy in initial screening of pattern sets HIV1 protease (wild type and mutants) of sub type B against the inhibitors Indinavir and Saquinavir. Genetic algorithm gives 80% Accuracy for Indinavir and 60% Accuracy for Saquinavir. Conclusion: Geno2pheno software uses machine learning analysis for subtypes of HIV with proper inhibitory values. If Molecular Mechanics is followed by Machine Learning with appropriate Inhibitory or effective concentration analysis, the validation of Genotyping will be more accurate than initial Geno2Pheno analysis. In future even the dynamics of the molecule will be analyzed with molecular mechanics and machine learning principles for various mutations of all FDA approved protease Inhibitors within the individual complex with the protease.