Machine Learning based Approach in Building QSAR Models for the Study of Asparagine Endopeptidase

Abstract

Recently the significance of chemometrics in Quantitative structure-activity relationship (QSAR) study has picked up the pace for dealing with a huge amount of information for discovering potential compounds from large databases. The process of selecting descriptors from a large number of physicochemical parameters has become very critical in building QSAR models. Machine learning techniques like evolutionary algorithms can be implemented for this purpose. As the Genetic algorithm (GA) can select optimal descriptors from a large set in minimum time with the help of its stochastic nature, in this work genetic algorithm(GA) has been used to develop QSAR models to predict the inhibition activity of Asparagine endopeptidase (AEP) which is an important enzyme having a key role in the treatment of breast cancer, colorectal cancer, gastric cancer, and tumors. Models were built using a set of 60 organic compounds showing excellent experimental interaction with the AEP enzyme. GA-based molecular docking has also been performed to validate the inhibition potency of the studied ligands. The ligand number 4 showed the least binding energy as −8.9. The ligand 4 formed 5 hydrogen bonds with the protein showing greater inhibition capability for AEP.