Illuminating Understudied Kinases and Facilitating Drug Discovery through Integrative Protein Kinase Resources and Machine Learning Methods

Protein kinases are the largest family of druggable proteins altered in various human cancers. Kinase-targeted drugs are emerging as a research target for personalized medicine because of the differential drug response shown in patients with kinase mutations. However, nearly 30% of the 545 human protein kinases remain understudied, and little is known about the systematic impact of drug-kinase mutation interactions. Although many biomedical resources with unique perspectives on the protein or drug landscape, generating new knowledge from these resources requires integrative and reproducible workflows. In this dissertation, I develop integrative resources and use machine learning methods for illuminating understudied kinases and investigating the contribution of drug-mutation interactions to drug response prediction. First, I develop a novel orthology inference method, the KinOrtho, to efficiently and accurately identify human kinase orthologs. Next, I integrate multiple resources using a federated query language. Last, I utilize integrative resources, introduce drug-mutation interaction terms to the prediction model, and use machine learning to predict kinase drug response in cancer cells. The findings presented in the dissertation facilitate the understanding of understudied kinase and shed insights into personalized medicine.