Descriptor-based profile analysis of kinase inhibitors to predict inhibitory activity and to grasp kinase selectivity

Hyejin Park, Kyeung Kyu Kim, Chang Hoon Kim, Jae Min Shin, Kyoung Tai No

Research output: Contribution to journalArticle

1 Citation (Scopus)


Protein kinases (PKs) are an important source of drug targets, especially in oncology. With 500 or more kinases in the human genome and only few kinase inhibitors approved, kinase inhibitor discovery is becoming more and more valuable. Because the discovery of kinase inhibitors with an increased selectivity is an important therapeutic concept, many researchers have been trying to address this issue with various methodologies. Although many attempts to predict the activity and selectivity of kinase inhibitors have been made, the issue of selectivity has not yet been resolved. Here, we studied kinase selectivity by generating predictive models and analyzing their descriptors by using kinase-profiling data. The 5-fold cross-validation accuracies for the 51 models were between 72.4% and 93.7% and the ROC values for all the 51 models were over 0.7. The phylogenetic tree based on the descriptor distance is quite different from that generated on the basis of sequence alignment.

Original languageEnglish
Pages (from-to)2680-2684
Number of pages5
JournalBulletin of the Korean Chemical Society
Issue number9
Publication statusPublished - 2013 Sep 20


All Science Journal Classification (ASJC) codes

  • Chemistry(all)

Cite this