Exploiting convergent evolution to derive a pan-cancer cisplatin sensitivity gene expression signature

Precision medicine offers remarkable potential for the treatment of cancer, but is largely focused on tumors that harbor actionable mutations. Gene expression signatures can expand the scope of precision medicine by predicting response to traditional (cytotoxic) chemotherapy agents without relying on changes in mutational status. We present a novel signature extraction method, inspired by the principle of convergent evolution, which states that tumors with disparate genetic backgrounds may evolve similar phenotypes independently. This evolutionary-informed method can be utilized to produce signatures predictive of response to over 200 chemotherapeutic drugs found in the Genomics of Drug Sensitivity in Cancer Database. Here, we demonstrate its use by extracting the Cisplatin Response Signature, CisSig, for use in predicting a common trait (sensitivity to cisplatin) across disparate tumor subtypes (epithelial-origin tumors). CisSig is predictive of cisplatin response within the cell lines and clinical trends in independent datasets of tumor samples. Finally, we demonstrate preliminary validation of CisSig for use in muscle-invasive cancer, predicting overall survival in patients who undergo cisplatin-containing chemotherapy. This novel methodology can be used to produce robust signatures for the prediction of traditional chemotherapeutic response, dramatically increasing the reach of personalized medicine in cancer. Translational RelevanceMost precision medicine research focuses on using targeted drugs on patients with known driver mutations, yet the majority of patients dont have actionable mutations. Using a novel signature extraction method, we produce the Cisplatin Response Signature (CisSig) to predict how well patients with epithelial-origin tumors will respond to cisplatin, a common cytotoxic chemotherapy. We show that expression of CisSig is correlated to clinical trends of cisplatin use in treatment guidelines using independent tumor databases. Then, we look at preliminary validation of CisSig for use in patients with muscle-invasive bladder cancer. Using two independent cohorts of pre-treatment tumor samples, we show that a CisSig-trained model is predictive of overall survival in patients who did receive cisplatin, but this signal is lost in patients who did not receive cisplatin-indicating that the model is predictive of therapeutic response, not simply prognosis.