Computational prediction of small molecules targeting Lassa fever drug target using quantitative structure activity relationship (QSAR) and random forest algorithm

Lassa fever, an endemic viral hemorrhagic fever in West Africa, is attributed to the Lassa virus as its causative agent, and this disease has led to the untimely death of many people in the affected areas. At present, the available treatment options for Lassa fever are limited and there is need for new drugs. This study aims to use computational tools to predict the efficacy of small molecules that can target the Lassa fever virus glycoprotein which is essential for viral entry into host cells. This study uses quantitative structure activity relationship (QSAR) to reduce the cost and time of preclinical evaluation of potential drugs. This study retrieves 7620 molecules that can inhibit Lassa virus glycoprotein from ChEMBL database and builds a regression model with random forest algorithm. Its performance was compared with other regression models by using lazy predict, and random forest performed better than most of the regression models. The coefficient of determination r2 are 0.93 and 0.56 for the training and test set and root mean square error (RMSE) of 0.32 and 0.77 for the training set and test set, respectively. In conclusion, the model satisfies the acceptable QSAR model. Key words: Quantitative structure-activity relationship, bioactivity, drug-likeness, drug target.