UAB Researcher Uses AI to Predict Cancer Therapy Response

Dr. Neil Pfister, an associate scientist at the UAB O’Neal Comprehensive Cancer Center and assistant professor in the Department of Radiation Oncology, is utilizing AI foundation models to predict how individual patients will respond to specific cancer therapies. This approach aims to replace general population averages with precision medicine based on a patient's unique biology.

To achieve this, Pfister and his collaborators developed CURE AI (Clinical trials Uncovering Real Efficacy Artificial Intelligence), a ground-up architecture created by Numenos. The system predicts whether a patient will benefit more from a new therapy compared to standard care by analyzing genetic information and medical record data, such as laboratory values, sourced from clinical trials. According to Dr. Pfister, the framework allows researchers to identify responders by distilling "real signal from millions of data points per patient," a feat he notes was previously impossible.

Overcoming Data Complexity

Predicting treatment response has long been hindered by the "curse of dimensionality," where the number of genomic markers often exceeds the number of patients in a study. While some researchers use simulated data to fill these gaps, Pfister argues that "faking" data does not improve medical care. Instead, his work focuses on using specialized model architectures to find genuine biological signals even within small patient cohorts. This methodology is detailed in the publication "A Deep Learning Framework for Causal Inference in Clinical Trial Design: The CURE AI Large Clinic genomic Foundation Model."

Impact on Rare Cancers

The research has already been applied to lung, renal, and pediatric cancers. By employing "pan-cancer" cross-indication selection, the AI can translate findings from common malignancies to rare ones. This capability is critical for patients with rare cancers, as it can significantly accelerate the timeline for accessing life-saving therapies that might otherwise take years to be validated for their specific condition.

Establishing Industry Standards

Beyond patient care, Pfister is working to ensure that clinical AI is objectively measured. He is collaborating with the FDA, the National Cancer Institute (NCI), and MLCommons to create standardized benchmarking datasets. These benchmarks will allow the medical community to evaluate the efficacy of various AI models using a consistent set of metrics.

The importance of this work was recently highlighted by Jorge Reis-Filho, AstraZeneca’s chief of AI for Science Innovation, during an NCI workshop titled "Advancing Cancer AI Benchmarks for Real-World Impact." As the field moves forward, the focus remains on integrating these standardized benchmarks into clinical practice to ensure AI-driven predictions are both safe and effective.