A Rough Set-Based Model of HIV-1 Reverse Transcriptase Resistome

A huge effort is being put into find a treatment for AIDS. HIV, the causative agent of AIDS has been studied in ever increasing detail to produce effective antiviral therapies. The high rate of replication and mutability of the virus leads to rapid drug-resistance in the virus. Efforts to overcome the AIDS pandemic would require drugs or drug regimens that can control the drug resistance in the virus.

Reverse transcriptase is one of the viral enzymes that is required for transcribing the RNA to DNA. This transcription is required for the viral genes to get integrated into the host genome. Only after integrating into the host genome, the virus can replicate and propagate. Drugs that inactivate this enzyme can be very effective in stopping the replication of the virus. However, the rapid emergence of drug resistance in the enzyme has made it difficult to treat AIDS with any single drug. Among 25 drugs currently used in HIV therapy, 12 attempt at inhibiting reverse transcriptase enzyme.

Drug resistance generally occurs due to a non-linear combination of mutations. Being able to predict if a drug will be effective against a particular mutant has been a useful tool in treatment. Further research has also given details about the mechanisms of drug resistance development and functionally important regions in the enzyme.

In this study, local phyiscochemical properties of a protein sequence where used to understand and predict drug resistance. Annotated data from the Stanford HIV database was used to classify the mutants into three groups labeled as susceptible, moderately resistant and resistant. The Monte Carlo feature selection was used to select the best features from a total of 7* 560 properties. The selected features where used to generate rules to classify the sequences into the correct class. The method was tried based on data available for different antivirals.

Evaluation of the results was done by 10 fold cross validation of the data. Performance of the classifier was assessed based on prediction accuracy and area under the curve. Analysis of the results for the sites responsible for the resistance where found to be in correlation with the known sites. New sites that could lead to resistance have also been predicted. Newly discovered sites seem to have the resistance effect by disturbing the complex network of hydrophobic and polar interactions responsible for stability of tertiary structure.