Structural Bioinformatics Approaches Toward Facilitating Drug Discovery

Structural Bioinformatics, as a branch of bioinformatics, is concerned about the expression, storage, access and analysis of the structural information at atomic and sub-cellular space size. In recent years, researchers have paid more attention to the applications of structural bioinformatics in drug development. Structural bioinformatics can be applied to almost all aspects of drug discovery, such as the evaluation of target assessment, virtual screening and lead optimization. In this paper, we reviewed the research framework of structural bioinformatics, given the practical databases and tools for structural bioinformatics and elaborated the important role of structural bioinformatics in drug discovery. My research mainly focused on the use of structural bioinformatics methods to promote drug discovery and development. The following three relatively independent works were carried out:1. Research on the new method of generating pharmacophore model for GPCRsG-protein coupled receptors (GPCRs) are the most prominent therapeutic target family characterized by seven conserved transmembrane (7TM) α-helical fold. With the aim to speed up the process of GPCR-based drug discovery, by studying the advantages and disadvantages of the existing methods and data mining from GPCR structures, we developed a new method, Pharm-Map-Pick, which can rapidly generate pharmacophore models for GPCR family. Our method mainly comprises three steps:constructing a library for key residues and pharmacophore features revealed from complex structures, mapping these pharmacophore features to a given GPCR and picking appropriate features to generate a final pharmacophore model. The results show that our method neither require a crystal structure nor ligand information. These models can accurately predict the binding mode of GPCR-ligand and has a high enrichment factor, thus can be directly used for virtual screening. We also built a world’s first database contains pharmacophore modes for all human GPCRs. This method and the resulting pharmacophore models will greatly contribute to the GPCR-based basic research and drug discovery.2. Genome-wide Prediction of Targets for AspirinBesides to anti-inflammatory, analgesic, and anti-pyretic properties, aspirin is used for the prevention of cardiovascular disease and various types of cancer. The multiple activities of aspirin cannot be attributed wholly to a single target and likely involve several molecular targets and pathways. In this study, we combined structural bioinformatics and systems biology approaches to predict potential targets and molecular pathways of aspirin in genome-wide. Firstly, using the tool of protein local structure detection and comparison together with molecular docking and binding free energy calculations, we identified21potential targets of aspirin. Further systems biology analysis of these targets, we found that aspirin participates in multiple molecular pathways, such as vascular endothelial growth factor, mitogen-activated protein kinase, JNK/p38, Fc epsilon RI and arachidonic acid metabolism. Based on these findings, we constructed a target of aspirin-cell effect interaction network. This network can explain a variety of medical efficacy of aspirin, which is used for anti-inflammatory as well as the prevention of cardiovascular disease and cancer. With the finish of this work, we have established a common platform to identify targets for small-molecule drugs. It can be used to find molecular targets for other small molecules.3. Identification and comparison of the druggable pocket at the interface of human PPIIn the last decade, PPIs have become the attractive molecular target for novel therapy. In this study, we identified1731druggable pocket at the interface of human PPI. Identification and comparison of these druggable pockets resulted in a lot of previously unknown interesting findings:1. Nearly half of the proteins have druggable pockets at the interface of human PPI;2. These proteins are enriched in the molecular pathways of various diseases;3. Compared to the protein binding interface, the frequencies of hydrophobic amino acids in the druggable pockets are increased;4. The partners of homo-PPI tend to have druggable pockets simultaneously. However, it is exactly the opposite for the partners of hetero-PPI;5. The druggable pocket similarity network built here can be used to guide the design of multi-target drugs. These findings have practical significance for discovery of the drug targeting protein interactions as well as the multi-target drugs.