Background: Computational biological modeling reveals many dysregulated signaling pathways responsible for hallmark behaviors of cancer and variable drug response in the population. A mechanistic model created for each patient using comprehensive genomic inputs can biosimulate downstream molecular effects of cell signaling and drugs for each patient’s personalized in silico virtual disease model. Singula TRI is designed to predict the outcome of specific therapies with a continuous TRI Score, 0 to 100, for each patient’s unique genomic network. Methods: TRI’s ability to predict Overall Survival (OS), Disease Free Survival (DFS) and Mandard – tumor regression grade (TRG) was prospectively evaluated in a retrospective cohort of gastroesophageal adenocarcinoma (GEA) from UK OCCAMS consortium. Random sampling stratified by clinical factors was used to split the data into independent training (N = 140) and validation (N = 131) subsets. Multivariate Cox Proportional Hazard (PH) and Proportional Odds models were used to predict survival and pathological response as a function of the pre-defined TRI and clinical thresholds compared with standard clinical factors. Results: 271 GEA patients were selected who had pre-chemo treated biopsies with 50x whole genome sequencing from the OCCAMS International Cancer Genome Consortium study. The median age was 65.6 years, 234 male and 30 female, with deceased median OS of 21.9 months and living of 49.9 months. There were 35 T2, 215 T3, 70 N0, 126 N1, 62 N2 and 266 M0. Patients were treated with physician prescribed chemotherapy treatments (PPT) according to UK clinical guidelines (SC). Biosimulation revealed that 99% of these tumors had deficiency in DNA repair genes. Other pathways included amplification of multi-drug resistance pumps, TP53 mutations and aberrations of the PI3K/AKT pathway genes. The table shows that TRI provides additional predictive information for OS and DFS beyond PPT and standard clinical factors. TRI was also predictive of TRG in univariate analysis. TRI scores were also generated for 82 alternate therapies for each patient enabling selection of optimal therapies with estimates of improvements in median OS and DFS compared to SC. Conclusions: In this cohort of patients, Cellworks Singula TRI was predictive of survival and TRG beyond clinical factors. These positive results suggest the utility of biosimulation-informed therapy selection to improve survival of GEA and validation in prospective clinical studies is warranted.

Adjusted for: ¹ age sex T-stage N-stage TRG & PPT; ² age sex & PPT.