AlphaFold Models Expand Drug Target Landscape and Drive Structure-Based Drug Discovery

The availability of new AlphaFold models significantly increases the number of potential targets in drug design and will drive progress towards more structure-based drug discovery. In this research, all reverse dockings were performed using Schr￶dinger's Glide software. Ligand compounds were processed using the LigPrep module to generate low-energy three-dimensional structures, while protein structures were processed using the Protein Preparation Wizard module. The binding sites on proteins were identified using the SiteMap module to generate a receptor grid and the Grid module to generate docking boxes with standard precision SP mode of Glide. Docking scores were based on a combination of van der Waals energy, Coulomb energy, lipophilic term, hydrogen bond term, metal bond term, reward, and punishment. This process was automated on a Linux system through Python script calls to each module.

Target proteins were screened based on scores of gscore < -8 kcal/mol against the drugbank target database to identify potential mechanisms of action. Additionally, proteins associated with side effects and toxicity were screened based on scores of gscore < -8 kcal/mol, using available data from the clinical chemistry literature. These proteins are involved in important cellular metabolic processes such as amino acid and nucleotide metabolism, glycolytic pathways, and the urea cycle, which may lead to adverse effects.