On May 16, Zheng Mingyue, a researcher of Shanghai Institute of Materia Medica, Chinese Academy of Sciences, used the meta learning technology combined with graph neural network to build KinomeMETA, an artificial intelligence (AI) platform that supports user customized kinase multi target activity screening, providing a powerful tool for discovery of new drug kinase targets and drug screening. Relevant research was published in Nucleic Acid Research.
Abnormal activation or overexpression of protein kinase is related to many diseases. Kinase targeting inhibitors are a class of valuable therapeutic drugs. It is necessary to systematically evaluate the multiple pharmacological effects of small molecule probes of kinase. Therefore, it is necessary to develop AI algorithms with wider application scenarios to screen potential kinase targets and multi target selective kinase small molecule probes on a large scale.
Based on previous research, researchers learned the action modes of active compounds in different kinases from more than 610000 kinase activity data, and built the KinemeMETA platform. The platform contains two core modules, of which the "customization" module allows users to use private data to build new kinase models or enhance the prediction ability of existing kinase models; The "Prediction" module can predict the probability of inhibitory activity of the compound against 661 wild and clinically relevant kinase mutants. The platform also provides selective analysis of compound kinases, evaluation of molecular properties and identification of similar inhibitors, assisting researchers in further research.
In addition, researchers screened compounds in the laboratory based on the "customized" function enhancement model of the platform, and found that the KinemeMETA platform can rapidly enhance the prediction ability through a few active compounds.
Overview of the construction and use process of the KinemeMETA platform. The picture is from Nucleic Acid Research
Relevant paper information: https://doi.org/10.1093/nar/gkae380
