Abstract

Parkinson's disease is one of the most common neurodegenerative disorders in elder people and the leucine-rich repeat kinase 2 (LRRK2) is a promising target for its pharmacological treatment. In this paper, QSAR models for identification of potential inhibitors of LRRK2 protein are designed by using an in house chemical library and several machine learning methods. The applied methodology works in two steps: first, several alternative subsets of molecular descriptors relevant for characterizing LRRK2 inhibitors are identified by a feature selection software tool; secondly, QSAR models are inferred by using these subsets and three different methods for supervised learning. The performance of all these QSAR models are assessed by traditional metrics and the best models are analyzed in statistical and physicochemical terms.