Challenge
Cancer is the leading cause of disease-related death for American children; roughly 15,000 are diagnosed each year, and those with relapsed, refractory, or rare cancers have few options. Unlike adult cancers, pediatric cancers are hundreds of rare, molecularly distinct diseases, producing small patient populations and fragmented data that defeat conventional approaches.
Realizing AI's potential for these children requires secure, interoperable, AI-ready data and advanced computing at national scale, and will change lives and save thousands of children.
AI Solution
Researchers will leverage AI to integrate with linked genomic, multi-omic, imaging, pathology, clinical, and survivorship data into multimodal models that learn across distributed sources such as electronic patient and family health records, claims, and other data, using privacy-preserving federated learning and trusted-access methods. This overcomes the limits of small populations, enabling robust models even for ultra-rare cancers and revealing relationships among molecular alterations, tumor biology, therapeutic targets, and outcomes. Autonomous experimentation will screen priority pediatric molecular targets at high throughput, and patient-level digital twins will predict treatment response, progression, and toxicity to guide precision care and adaptive trial design. AI methods will be used to improve clinical trial design and execution, and AI models will advance the development of diagnostic, prognostic, and therapeutic biomarkers in pediatric cancer.
Justification
Executive Order 14355, sustained federal investment in a national pediatric cancer data ecosystem, and established laboratory collaborations provide the scientific and organizational foundation for this Challenge. The Department of Health and Human Services contributes the integrated pediatric cancer data ecosystem, the national network of designated cancer centers through which models can be trained and deployed, and clinical and regulatory expertise while protecting patient privacy; the Department of Energy contributes the computational scale required to train multimodal foundation models across hundreds of rare subtypes. Neither capability delivers impact without the other.
National Impact
This Challenge will accelerate biomarker discovery, target identification, precision medicine, and trial optimization for every child with cancer, while improving survivorship. It will establish the first national learning system that continuously integrates multimodal pediatric cancer data: a privacy-preserving model that is additionally usable across rare adult cancers and other diseases with small samples and distributed data.
Challenge
This was last updated .