Kinases are firmly established drug targets in cancer. There are currently 44 FDA-approved kinase drugs and hundreds of compounds are in clinical development. However, less than 10% of the Kinome is currently targeted and a large proportion is considered understudied by the NIH Illuminating the Druggable Genome Program (https://druggablegenome.net/). No small molecule inhibitors are known for these “dark” proteins, yet many may be opportune novel cancer targets. We developed a computational pipeline to identify and prioritize understudied kinases as cancer drug targets. We analyzed the complete set of tumors in The Cancer Genome Atlas (TCGA). For 33 different cancers, we performed differential expression analysis and identified 39 dark kinases that exhibit significant upregulation in at least four types. Using co-expression analysis we built functional networks prioritizing drug targets. To identify small molecules that reverse their expression levels, we leveraged transcriptional response signatures obtained from dozens of human cancer cell lines exposed to tens of thousands of small molecules from the Library of Integrated Network-based Cellular Signatures (LINCS). To identify small molecules that directly bind to and inhibit dark kinases, we have combined an advanced AI (artificial intelligence) model trained on activity data from across the Kinome with structure-based simulations. Using the computational pipeline, we identified the dark Ca2+/Calmodulin dependent kinase PNCK as the most differentially overexpressed kinase in kidney cancer patients. Our analyses have demonstrated a statistically significant correlation between PNCK mRNA levels and various clinical and pathological outcomes, including histologic grade, clinical staging, and overall survival. We have confirmed high levels of PNCK expression in 5 renal cell carcinoma cell lines (Caki-1, ACHN, 786-O, A704 and A498). Knockdown and overexpression studies have suggested PNCK and the CaMK pathway may contribute to cellular proliferation and cell cycle progression. We have applied our AI-based screening pipeline to a library of >20 million commercially available compounds and confirmed three PNCK inhibiting chemotypes. In summary, using a novel computational pipeline, we have identified and experimentally validated PNCK as a prospective novel drug target in an understudied pathway that is highly upregulated in kidney cancer. We identified first in class small molecules that target this previously dark kinase as prospective starting points for optimization into a clinical candidate.
Read more . . .