No abstract

Human immunodeficiency virus (HIV) is a virus that infects an immune cell and makes the patient more susceptible to infections and other diseases. HIV is also a factor that leads to acquired immune deficiency syndrome (AIDS) disease. The active target that is usually used in the treatment of HIV is HIV-1 protease. Combining HIV-1 protease inhibitors and reverse-transcriptase inhibitors in highly active antiretroviral therapy (HAART) is typically used to treat this virus. However, this treatment can only reduce the viral load, restore some parts of the immune system, and failed to overcome the drug resistance. This study aimed to build a QSAR model for predicting HIV-1 protease inhibitor activity using the gravitational search algorithm-neural network (GSA-NN) method. The GSA method is used to select molecular descriptors, while NN was used to develop the prediction model. The improvement of model performance was found after performing the hyperparameter tuning procedure. The validation results show that model 3, containing seven descriptors, shows the best performance indicated by the coefficient of determination (r2) and cross-validation coefficient of determination (Q2) values. We found that the value of r2 for train and test data are 0.84 and 0.82, respectively, and the value of Q2 is 0.81.

No abstract

No abstract