Development of Practical Artificial Intelligence System for Drug Discovery and Its Application to Activity Prediction of Small Molecule Protein-Protein Interaction Modulators

Abstract

Degree of attention to protein-protein interaction (PPI) are increasing under circumstances of the exhaustion of conventional drug targets such as enzymes, receptors and channels whose ligands are non-protein. However, identification of potent small molecule PPI modulators are still challenging for pharmaceutical industries because in-house chemical libraries of pharmaceutical companies do not include strong and specific binders to PPI targets, and it is difficult to identify active compounds by high-throughput screening (HTS). On the other hand, to improve low productivity of drug research and developments (R&D), which is one of the most important longstanding issues for pharmaceutical industries, application of artificial intelligence (AI) to many stages of drug R&D is being tried by large number of pharmaceutical companies, academic institutions and biotech companies. However, a part of researchers of pharmaceutical companies also have opinions with a skeptic tone for the contribution of AI to the improvement of low productivity of drug R&D. Based on these situations, we review current approaches of application of AI to drug discovery researches and introduce Interprotein’s approach to identify potent small molecule PPI modulators with an AI-based activity prediction system.