Machine learning empowered drug discovery

Abstract

Traditional drug discovery strategies include lead molecule identification, lead optimization, preclinical studies and clinical trials. The pharmaceutical and biotechnology research and development (R&D) department spends more than 10 years and $1 billion to bring the molecule to market successfully. About 90% of drug candidates fail in the drug development due to safety and efficacy issues. The lack of technologies is the main limitation for identifying potential candidates from the available chemical space (>1060 molecules). De Novo design methods explore chemical space through pharmacophore (ligand-based), and docking (structure-based) approaches. Structure-based drug discovery approaches use the insights gained from biological data of target structures. Schrödinger, AutoDock and Biovia (Accelrys) pioneered the development of structure-based tools to improve drug discovery. Libraries of molecules can be screened for their target suitability, known as virtual screening. The structure-based drug discovery approach uses the three-dimensional (3D) details of the target structure and explains the intermolecular interactions (biophysical simulations). Ligand-based drug discovery approaches are based. Read more.........