The University of Texas MD Anderson Cancer Center has received a USD 3 million NIH cooperative agreement to validate predictive tumor models — combining in vitro organoids and in silico deep learning tools — directly within ongoing colorectal cancer clinical trials, a step toward making these models clinically actionable rather than purely experimental.
The OPTIMA (Oncology Precision Trials using Individualized Model Avatars) program, led by principal investigators Anil Korkut, Ken Chen, and Xiling Shen, will embed three combinatorial New Approach Methodologies (NAMs) — micro-organospheres, the DL-REFLECT deep learning model, and VirtualCellTherapy — into three clinical trials already running at the institution. The co-funding structure, split between the National Cancer Institute (USD 1.5 million) and the NIH Office of the Director (USD 1.51 million), reflects cross-institute interest in advancing non-animal predictive models for oncology.
The central challenge the project addresses is well-established: organoid and computational models have demonstrated preclinical promise in predicting patient drug responses but have not been standardized or prospectively validated at a scale sufficient to influence treatment decisions. By embedding model evaluation within active trials, the team aims to generate the clinical evidence needed for regulatory-grade assay development, with the MD Anderson Emerging Diagnostics Center targeting College of American Pathologists accreditation for the resulting laboratory developed tests.
Three clinical applications are prioritized: refining second-line treatment selection for metastatic colorectal cancer, identifying patients with minimal residual disease who currently have no established treatment guidelines, and improving patient selection for CAR-NK cell therapy in solid tumors — an area where cell therapies have struggled to replicate the efficacy seen in hematologic malignancies.