About KinasePhos 3.0

In 2005, our group developed KinasePhos 1.0 for identifying protein kinase-specific phosphorylation sites. The tool constructed models from the kinase-specific groups of the phosphorylation sites based on the profile hidden Markov model (HMM). Then, support vector machines (SVM) with the protein sequence profile and protein coupling pattern was applied to update the tool to Kinasephos 2.0. Due to the rapid development of phosphorylation-related research, the datasets used for training are constantly expanding. As an expansion of KinasePhos 1.0 and 2.0, in this study, we introduce KinasePhos 3.0 to improve the performance of kinase-specific phosphorylation site prediction. Experimentally verified kinase-specific phosphorylation data were collected from PhosphoSitePlus, UniProt, GPS5.0, and Phospho.ELM. In total, 41,421 experimentally verified kinase-specific phosphorylation sites and 1,380 unique kinases were identified, including 753 kinases with known classification information based on KinBase and the other 627 kinases annotated by building an evolutionary tree. Based on this kinase classification, 771 machine learning-based kinase-specific phosphorylation site prediction models were built at kinase group, kinase family, and individual kinase levels, with at least 15 experimentally verified substrate sites considered in each model.

Citing publications