Abstract A070: Clinical application of functional precision medicine for pancreatic cancer enables personalized treatment prioritization and identification of RAS-targeted combinations
Edward Kai-Hua. Chow, Masturah Mohamed Abdul Rashid, Lydia Ogrodzinski, Jhin Jieh Lim, Sharon Chan, Ralph GarippaAbstract
Background:
Treatment of advanced pancreatic cancer relies on empirically selected combinations yielding response rates of 20 to 40%. Although KRAS mutations are present in over 90% of these cases, responses to KRAS-targeted therapies vary, highlighting the need for functional approaches to identify optimal therapies. Optim.AI™, a functional precision medicine platform using small data AI, has previously demonstrated clinical utility for hematological cancers and sarcoma. In this study, we evaluated the real-world implementation of Optim.AI™ in pancreatic cancer within a CAP/CLIA laboratory while demonstrating its analytical robustness and ability to identify effective RAS-targeted treatment combinations.
Methods:
Tumor cells were isolated from solid tissues or body fluids of pancreatic cancer samples. Short-term patient-derived organoids were formed before combinatorial treatment with a customized 12-drug panel of chemotherapeutic and targeted agents. Based on this small dataset, Optim.AI™ predicts all possible combinations and ranks them by sensitivity for individual patients. To evaluate inter-laboratory robustness, proficiency testing on pancreatic cancer cell lines, AsPC-1 and MIA PaCa-2, was independently tested at two laboratories. Assay quality parameter, Z' factor and Optim.AI™-predicted single-drug rankings between the two laboratories were compared.
Results:
Approximately 77% of the samples processed had sufficient viable cells to proceed with Optim.AI™ testing, where reports were successfully generated for 92% of these samples. The mean turnaround time from sample receipt to dissemination of clinical reports was seven business days. Aggregated analyses of Optim.AI™ reports demonstrate differential sensitivities towards standard of care options, with incremental decrease in sensitivity observed for both off-label and novel treatments. Optim.AI™ demonstrated a negative concordance of 75%, accurately predicting resistance to prior or ongoing therapies. Inter-laboratory proficiency testing met all pre-defined acceptance criteria, with both laboratories achieving Z' factors >0.5 and concordant identification of three of the four highest-ranked single agents across both cell lines. Optim.AI™ analyses from these preclinical models also identified Gemcitabine-containing RAS-targeted combinations being consistently top-ranked, a trend that was similarly observed in a KRAS G12V patient-derived sample. Synergistic interactions were also observed between RAS- and EGFR- inhibitors.
Conclusions:
This study demonstrates the successful clinical implementation of Optim.AI™ for pancreatic cancer within a CAP/CLIA laboratory, providing actionable treatment recommendations within seven business days. High concordance with observed clinical resistance supports its potential to improve treatment selection, while functional drug profiling identified promising RAS-based combinations that may accelerate development of combinatorial therapies. Further validation in larger cohorts representing diverse KRAS mutational backgrounds is warranted.
Citation Format:
Edward Kai-Hua. Chow, Masturah Mohamed Abdul Rashid, Lydia Ogrodzinski, Jhin Jieh Lim, Sharon Chan, Ralph Garippa. Clinical application of functional precision medicine for pancreatic cancer enables personalized treatment prioritization and identification of RAS-targeted combinations [abstract]. In: Proceedings of the AACR Conference on Pancreatic Cancer: New Frontiers in Biology and Therapeutic Development; 2026 Sep 25-28; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(18_Suppl_2):Abstract nr A070.