Quantiative structure-activity relationship analysis of 1-[(2-Hydroxyethoxy)methyl]-6-(phenylthio)thymine (HEPT) analogues with the use of random forest

A quantitative structure-activity relationship (QSAR) analysis on anti-HIV-1 analogue series of 1-[(2-Hydroxyethoxy)methyl]-6-(phenylthio)thymine (HEPT) has been studied. 132 HEPT analogues were gathered along with the experimental data of log(1/EC50). Molecular descriptors were generated in order to reveal the numerical properties of HEPT analogues. Due to the immense amount of descriptors available, a random forest technique was used to indicate potential descriptors and make a prediction on the values of log(1/EC50). Out of 2,976 generated trees, the best tree has selected six descriptors with the highest r2 of 0.929 between the predicted log(1/EC50) and the actual log(1/EC50). Results indicated that descriptors involving the hydrophobicity of the functional group provide a major importance in relation to the effectiveness of HEPT's inhibition towards the human immunodeficiency virus type 1 reverse transcriptase.