Abstract LB-319: Predicting clinical outcomes in cancer patients based on a novel computational biology method of mapping the cancer mutanome

Abstract

Abstract Background: Despite three FDA approved standard of care (SOC) therapeutics, 60% of patients with myelodysplastic syndromes (MDS) fail to achieve hematologic improvement and nearly all patients die of refractory disease. The ability to predict patient responses to SOC therapy would improve treatment efficacy, avoid treatment-related adverse events and reduce health care costs. Ideally, a patient's prediction would be based on their cancer mutanome. Methods: A prospective clinical trial was initiated in June 2015 to (1) predict patient response to SOC, and (2) identify a personalized therapy of FDA approved drugs that target the patient's unique dysregulated pathways. Currently, 50 patients have been recruited to iCare For Cancer Patients (NCT02435550). Bone marrow and/or peripheral blood are collected from each patient and utilized for whole-exome sequencing (Agilent OneSeq) and copy number variation (Agilent array CGH) analysis. Cytogenetics, FISH, gene mutations, and CNV data are inputted into a new computational biology computer program, which generates a patient-specific map of multiple and intersecting protein networks. Mathematical modeling is then used to simulate cancer cell proliferation, viability and apoptosis. The patient's simulation results are then used to determine if the cancer is sensitive or resistant to SOC. Additionally, each patient's cancer model is used to digitally screen for FDA approved drugs with therapeutic potential. A database of de-identified was created that includes genomic information from each patient with thorough clinical annotation. Results: Genomic profiling was completed in 49/50 patients, indicating high feasibility of this functional genomics workflow. To date, 18/50 patients are evaluable. Of 18 patients, 7 have clinical response information available. 2/7 patients were predicted to respond to therapy, and 2/2 had a clinical response to therapy, yielding a 100% positive predictive value. 5/7 patients were predicted to not respond to therapy, and 5/5 had no clinical response to therapy, yielding a 100% positive predictive value. 7/7 patient outcomes (100%) were correctly matched to their predicted response, showing a sensitivity and specificity of 100%. Conclusion: These results demonstrate the feasibility of a new precision oncology method of functional genomics and computational biology. Preliminary results in this prospective study support high accuracy of prediction in retrospective studies by our group. These data also support an upcoming randomized phase II clinical trial of physician's choice versus prediction model-informed treatment in patients with refractory/relapsed cancer. Citation Format: Leylah Drusbosky, Regina Martuscello, Taher Abbasi, Shireen Vali, Christopher R. Cogle. Predicting clinical outcomes in cancer patients based on a novel computational biology method of mapping the cancer mutanome. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr LB-319.