Pre-clinical validation of an RNA-based precision oncology platform for patient-therapy alignment in a diverse set of human malignancies resistant to standard treatments

Abstract

Predicting tumor sensitivity to antineoplastics remains an elusive challenge, with no methods demonstrating predictive power. Joint analysis of tumors--from patients with distinct malignancies who had progressed on multiple lines of therapy--and drug perturbation transcriptional profiles predicted sensitivity to 28 of 350 drugs, 26 of which (93%) were confirmed in low-passage, patient-derived xenograft (PDX) models. Drugs were prioritized based on their ability to either invert the activity of individual Master Regulator proteins, with available high-affinity inhibitors, or of the modules they comprise (Tumor-Checkpoints), based on de novo mechanism of action analysis. Of 138 PDX mice enrolled in 16 single and 18 multi-protein treatment arms, a disease control rate (DCR) of 68% and 91 %, and an objective response rate (ORR) of 12% and 17%, were achieved respectively, compared to 6% and 0% in the negative controls arm, with multi-protein drugs achieving significantly more durable responses. Thus, these approaches may effectively complement and expand current precision oncology approaches, as also illustrated by a case study.