| Summary: | <p>A primary aim of drug discovery is to find novel molecules that are active against a target of therapeutic relevance. With high attrition rates and the increasing costs of each stage of drug development, there is an emphasis on making the right decisions in the early stages of drug discovery. Methods are needed that will aid selection of compounds as part of the hit-to-lead process that will avoid subjective decisionmaking and make use of the recent increase in experimental data available at the start of the drug discovery process.</p> <p>I have developed the CRANkS algorithm to prioritise candidate compounds based on how novel the compounds are compared to known inhibitors of the target of interest. A prospective compound is compared to known binders of the target in terms of its predicted interactions with the protein, placement in the binding site, and chemical structure. Molecules are then scored based on the overlap of conformations with grids generated from protein-ligand X-ray crystal structures. The grids capture the spatial distribution of the chemistry of the ligands and proteinligand interactions.</p> <p>By using the calculated novelty scores, I hypothesised that the algorithm will prioritise both compounds that are active, but also compounds that are novel scaffolds. Scaffold-hopping is of particular importance in drug discovery for a better exploration of the activity space in order to find the most amenable lead compound. Using multi-objective optimisation, I found that the CRANkS scores could be used to select novel active scaffolds from datasets of active and inactive compounds. </p> <p>Finally, I have investigated “active-guided” docking by combining the grids generated from active ligands, developed as part of CRANkS, with AutoDock – “AutoCRANkS”. I found an improvement in discriminating between active and inactive compounds when using AutoCRANkS compared to AutoDock, indicating the promise of using protein-ligand structural data to guide docking.</p>
|
|---|