Computational methods for the quantitative prediction of drug properties

Abstract

Recently, there have been many computational approaches to save the expensive cost and time paid for a long drug discovery process. The computational methods usually tried to predict important drug properties such as drug toxicity in advance in order to select more qualified drug candidates. However, in spite of the importance, there is still plenty of room to develop efficient prediction-tools. Accordingly, in this study, I tried to develop the computational methods for the quantitative prediction of three important drug properties: multi-modal binding propensity, drug toxicity, and drug targets. At first, the basal studies which helped to question the existence of multi-modal binding molecules would be introduced. From the studies, I tried to answer the following questions: how much ligand and binding site are associated with protein function, and how ligands themselves are related to each other in terms of binding site because the systematic study about ligand structure, binding site, target function might reveal the important information about drug binding mechanism. For this, binding similarity network of ligand was presented, with the network analysis. The results showed that the ligand binding site and function were closely related (conservation-ratio 81%), and showed strong conservative tendency to function in line with ligand structure similarity. In addition, hubs ligands were also investigated in terms of their functional role in the binding similarity network. The ligand network analysis also suggested that some ligands showed promiscuous binding property which made questions whether the promiscuous binding property could be predicted and there existed the features determining the behavior. For this, the binding sites for all ligands in PDB were clustered by binding site similarity and the ligands that bind to many dissimilar binding sites were defined as multi-modal binding ligands. In addition, the quantified importance measures for global and l...