Project Overview

Selecting the right treatment, for the right patients, at the right time

Cancer is a complex, evolving disease in which each patient’s tumor presents a unique and multifaceted profile. Current approaches largely overlook this complexity, leading to suboptimal choice of therapy and frequent drug resistance. Artificial intelligence (AI) and deep learning have the power to transform cancer treatment by turning the vast, complex data landscape into actionable insights. The Dynamic Digital Tumors team seeks to substantially increase the pace and power of model development in precision oncology, resulting in a general-purpose drug response prediction engine that dynamically tracks tumor evolution and resistance to enable clinicians to select the right treatment for the right patients at the right time.

Together, we are developing, validating, and deploying an array of turnkey AI technologies:

**Check out our recently published Commentary in Cancer Cell to learn more about how the ADAPT Program aims to revolutionize cancer treatment** [Link] [PDF]

Leveraging Cell Maps for Interpretable AI

Our work builds on foundational advances from the NIH Bridge2AI Cell Maps for AI (CM4AI) initiative, which is generating comprehensive, multiscale maps of cellular organization and function in multiple disease contexts, including cancer. These maps integrate diverse data types to represent how genes, proteins, and pathways interact across biological scales—providing a critical foundation for understanding tumor behavior.

Within the Dynamic Digital Tumors project, we are leveraging these cell maps to develop biologically informed and interpretable AI models, including visible neural networks (VNNs) and related approaches. By embedding prior knowledge of molecular pathways directly into model architecture, these methods move beyond traditional “black box” AI, enabling predictions that are both accurate and mechanistically explainable.

This integration allows us to connect patient-specific tumor data to underlying biological processes, improving our ability to identify drivers of drug response and resistance—and ultimately supporting more transparent and actionable precision oncology.

CM4AI.org