Short-term HIV therapy response prediction using sequence information

HIV causes 3 million deaths annually. Advancements in medical sciences have enabled us to manage the infection with drug therapies in the recent years. However, HIV-1 has high viral variability and is likely to evolve resistance against these drugs, considering a high correlation between the evolutionary rate and disease progression. It is important to understand the genetic blueprint of the virus and the marker mutations that are linked with disease progression under treatment. For a data set containing clinical patient data at the be-ginning of the therapy, and the reverse transcriptase (RT) and protease (PR) nucleotide sequences of HIV-1 virus, we developed an algorithm to extract a number of features and predict the short term progression of the disease with re-sponse to the therapy and find the important positions within the sequences. The algorithm resulted in around 30 positions that can predict the disease progression with AUC of 0.824 and accuracy of 0.737, better than the standard methods and comparable to the best methods available on such a data.