| 研究実績の概要 |
In order to precisely and efficiently predict the binding potentials of test compounds against proteins involved in a molecular pathway, we have developed a network pharmacology-based prediction pipeline. It is mainly composed of a high-precision scoring function for molecular simulation with a well-designed machine learning model. This pipeline enables researchers to predictively screen a large number of small molecules over a complex molecular pathway, allowing comprehensively identifying the on-/off-targets. We have also developed a publicly accessible website sharing the screening facility to researchers, dedicating our achievements to the community of drug discovery.
For prediction validation, we tested our method using PDBbind dataset, containing about three thousand protein-ligand complexes. By assessing the correlations between the prediction scores and the experimental binding affinities, it shown a good performance in predicting the binding potentials. The correlations have been improved to >0.8. Additionally, we predicted the selectivity of various kinase inhibitors by comparing with known bioassay results, showing a good consistency.
The relevant research results have been published on high-impact journals, including Nucleic Acids Research, Scientific Reports and IEEE. We have also applied it to several joined projects helping collaborators, including those in Systems Biology Institute (SBI, Tokyo) and The University of Tokyo (IMSUT), to identify druggable molecules.
|
