Biology- and Physics-Informed Explainable AI Across the Lifespan

When beginning your next investigator-initiated application, consider the following NIH highlighted topic. The area of science described below is of interest to the listed NIH Institutes, Centers, and Offices (ICOs). This is not a notice of funding opportunity (NOFO).

Apply through an appropriate NIH Parent Funding Announcement or another broad NIH opportunity available on Grants.gov. Learn how to interpret and use Highlighted Topics.

Topic Description

Post Date: June 2, 2026

Expiration Date: June 2, 2028

Purpose
This highlighted topic encourages innovative research applying explainable artificial intelligence (XAI) across the lifespan, with a strong emphasis on mechanistically grounded, physics-based or biology-based explainability. The goal is to advance interpretable artificial intelligence (AI) approaches that enhance understanding of biological and health-related mechanisms, heterogeneity, and temporal dynamics across the lifespan, while supporting robust prediction of health and disease risks across the life course, including interpretation of past data if available.

Background
Health and functional trajectories across the lifespan are complex, dynamic, and heterogeneous processes shaped by interacting biological, physical, environmental, and social factors across multiple scales and over time. Although AI and machine learning methods have demonstrated utility in predicting health-related outcomes, many existing approaches function as black boxes, limiting mechanistic insight, biological interpretability, and translational relevance. There is a critical need for AI methods that are explainable, grounded in established principles of biology or physics, and capable of modeling within-individual change and between-individual variability over time.

Research Objectives
Applications in this topic area should develop or apply XAI methods that move beyond purely statistical prediction to provide interpretable representations of dynamic biological processes across the lifespan. Of particular interest are approaches that incorporate mechanistic constraints, causal or dynamical systems models, domain-informed priors, or hybrid models integrating data-driven learning with first-principles knowledge. Applications should develop explainable AI grounded in biology- and physics-informed principles, ensuring reproducibility and enabling mechanistic understanding, hypothesis generation, and translational relevance.

Participating ICOs

National Institute on Aging (NIA)

Examples of research in this area include, but are not limited to:

  • Longitudinal Aging Trajectories: XAI approaches that model within-person aging dynamics, identify critical transitions, checkpoints, or inflection points, and relate these to underlying mechanisms.
  • Heterogeneity of Aging: Methods that characterize inter-individual variability, identify aging subtypes or divergent trajectories, and provide interpretable explanations of resilience, vulnerability, or differential responses.
  • Prediction and Temporal Interpretation: Interpretable models that predict future health, functional decline, or disease risk, while also enabling retrospective interpretation of prior exposures or health states.
  • Multi-Scale and Multi-Modal Integration: Explainable integration of molecular, cellular, physiological, behavioral, environmental, and clinical data across time.
ICO Scientific Contact:
Leonid Tsap, Ph.D.
[email protected]

For technical issues E-mail OER Webmaster