Abstract
BackgroundProtein kinases are a large family of druggable proteins that are genomically and proteomically altered in many human cancers. Kinase-targeted drugs are emerging as promising avenues for personalized medicine because of the differential response shown by altered kinases to drug treatment in patients and cell-based assays. However, an incomplete understanding of the relationships connecting genome, proteome and drug sensitivity profiles present a major bottleneck in targeting kinases for personalized medicine.ResultsIn this study, we propose a multi-component Quantitative Structure–Mutation–Activity Relationship Tests (QSMART) model and neural networks framework for providing explainable models of protein kinase inhibition and drug response (hbox {IC}_{50}) profiles in cell lines. Using non-small cell lung cancer as a case study, we show that interaction terms that capture associations between drugs, pathways, and mutant kinases quantitatively contribute to the response of two EGFR inhibitors (afatinib and lapatinib). In particular, protein–protein interactions associated with the JNK apoptotic pathway, associations between lung development and axon extension, and interaction terms connecting drug substructures and the volume/charge of mutant residues at specific structural locations contribute significantly to the observed hbox {IC}_{50} values in cell-based assays.ConclusionsBy integrating multi-omics data in the QSMART model, we not only predict drug responses in cancer cell lines with high accuracy but also identify features and explainable interaction terms contributing to the accuracy. Although we have tested our multi-component explainable framework on protein kinase inhibitors, it can be extended across the proteome to investigate the complex relationships connecting genotypes and drug sensitivity profiles.
Highlights
Protein kinases are a large family of druggable proteins that are genomi‐ cally and proteomically altered in many human cancers
Performance of Quantitative Structure–Mutation–Activity Relationship Tests (QSMART) is comparable to deep neural networks (DNN) The QSMART model with neural networks predicts protein kinase inhibitor (PKI) responses in 23 cancer types with accuracies ranging from R2 = 0.805 to 0.881
Compared with commonly used machine learning models and a state-of-the-art DNN model, multiscale convolutional attentive (MCA) [36], the QSMART model with neural networks shows higher or comparable performances of predicting PKI response for 23 cancer types based on 10-fold cross-validation (Fig. 2b and Table 2)
Summary
Protein kinases are a large family of druggable proteins that are genomi‐ cally and proteomically altered in many human cancers. Kinase-targeted drugs are emerging as promising avenues for personalized medicine because of the differential response shown by altered kinases to drug treatment in patients and cell-based assays. An incomplete understanding of the relationships connecting genome, proteome and drug sensitivity profiles present a major bottleneck in targeting kinases for personalized medicine. A major contributing factor in drug resistance [2], as well as drug sensitivity [3], is the accumulation of mutations in oncogenic proteins such as protein kinases, which are primary targets for cancer drugs [4]. Mutations in protein kinases can have varying impacts on drug sensitivity depending on the structural location of mutations. As mutations impact the efficacy of different cancer drugs, there is a need to incorporate structural knowledge in drug response prediction methods
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
