THE ADVANCEMENTS AND CHALLENGES OF MACHINE LEARNING IN CHEMINFORMATICS

Abstract

Machine Learning (ML) has become one of the powerful techniques in Cheminformatics, also known as Computational Chemistry. It has been applied to various problems in the field. For instance, in Chemistry, machine learning is used in drug discovery, toxicity prediction, and materials design. In this paper, we aim to provide a general survey of ML in Cheminformatics. We begin by discussing the fundamental concepts of ML and then explore different types of ML algorithms that have been applied to Cheminformatics problems. This provides researchers and practitioners in the field of Cheminformatics with a comprehensive understanding of the application of computational techniques and methods. Additionally, we present some challenges and opportunities for further research. The final part of paper showcases a case study on activity prediction based on a dataset containing screening assays performed by the Burnham Center for Chemical Genomics, targeting the inhibition of VCAM-1 cell surface expression induced by the TNFa gene. This dataset exhibits a significant imbalance between active and inactive compounds, making it an interesting sample for experimentation. The results indicate that the selected classification model is relatively appropriate based on the AUC metric and F1-Score