Structure-guided machine learning prediction of drug resistance mutations in Abelson 1 kinase

• DOI: 10.1016/j.csbj.2021.09.016
• Publication Date: 2021-09-16T15:54:38Z
• Authors: Yunzhuo Zhou; Stephanie Portelli; Megan Pat; Carlos H.M. Rodri...; Thanh-Binh Nguyen; Douglas E.V. Pires; David B. Ascher
• Keywords: ABL

Kinases play crucial roles in cellular signalling and biological processes with their dysregulation associated diseases, including cancers. Kinase inhibitors, most notably those targeting ABeLson 1 (ABL1) kinase chronic myeloid leukemia, have had a significant impact on cancer survival, yet emergence of resistance mutations can reduce effectiveness, leading to therapeutic failure. Limited effort, however, has been devoted developing tools accurately identify ABL1 mutations, as well providing insights into molecular mechanisms. Here we investigated the structural basis modulating binding affinity eight FDA-approved drugs. We found impair type I II inhibitors differently used this insight developed novel web-based diagnostic tool, SUSPECT-ABL, pre-emptively predict profiles free-energy changes (ΔΔG) all possible against different modes. Resistance were successfully identified, achieving Matthew's Correlation Coefficient up 0.73 resulting change ligand Pearson's correlation 0.77, performances consistent across non-redundant blind tests. Through an silico saturation mutagenesis, our tool identified possibly emerging which offers opportunities for vivo experimental validation. believe SUSPECT-ABL will be important not just improving precision medicine efforts, but facilitating development next-generation that are less prone resistance. made freely available at http://biosig.unimelb.edu.au/suspect_abl/.