Computational prediction of anti HIV-1 peptides and in vitro evaluation of anti HIV-1 activity of HIV-1 P24-derived peptides.

The world is entering the third decade of the acquired immunodeficiency syndrome (AIDS) pandemic. The primary cause of the disease has known to be human immunodeficiency virus type I (HIV-1). Recently, peptides are shown to have high potency as drugs in the treatment of AIDS. Therefore, in the present study, we have developed a method to predict anti-HIV-1 peptides using support vector machine (SVM) as a powerful machine learning algorithm. Peptide descriptors were represented based on the concept of Chou's pseudo-amino acid composition (PseAAC). HIV-1 P24-derived peptides were examined to predict anti-HIV-1 activity among them. The efficacy of the prediction was then validated in vitro. The mutagenic effect of validated anti-HIV-1 peptides was further investigated by the Ames test. Computational classification using SVM showed the accuracy and sensitivity of 96.76% and 98.1%, respectively. Based on SVM classification algorithm, 3 out of 22 P24-derived peptides were predicted to be anti-HIV-1, while the rest were estimated to be inactive. HIV-1 replication was inhibited by the three predicted anti-HIV-1 peptides as revealed in vitro, while the results of the same test on two of non-anti-HIV-1 peptides showed complete inactivity. The three anti-HIV-1 peptides were shown to be not mutagenic because of the Ames test results. These data suggest that the proposed computational method is highly efficient for predicting the anti-HIV-1 activity of any unknown peptide having only its amino acid sequence. Moreover, further experimental studies can be performed on the mentioned peptides, which may lead to new anti-HIV-1 peptide therapeutics candidates.