ADIA Lab Health Science RFP – 2025:

Award Announcement

During 2025, ADIA Lab’s inaugural Health Science Request for Proposals attracted nearly 200 registered researchers from more than 20 countries, signaling the Lab’s emergence as a global convener of impactful, AI-driven health research. Following an open call, approximately 30 teams were invited to submit full proposals across four targeted thrust areas: real-world health data; multimodal AI for health science; AI for prevention and management of cardiometabolic diseases; and AI and computational tools for healthspan.

All full proposals underwent a rigorous single-blind peer-review process conducted by an expert reviewer panel drawn from leading academic medical centers and AI research institutions worldwide, followed by an independent evaluation by the ADIA Lab Advisory Board. ADIA Lab publicly announced three winning projects on the opening day of the 2025 ADIA Lab Symposium on October 28, 2025.

About the Awards

Each awarded project is funded with up to US $300,000 (inclusive of direct and indirect costs) over a two-year performance period, with the principal investigators joining ADIA Lab as new ADIA Lab Fellows for the duration of the award.

The Awarded Projects

The principal investigators below will work with ADIA Lab as ADIA Lab Fellows over the next two years on the following projects:

1. AgentiCDS: A Multimodal Agentic AI System for In-Hospital Clinical Decision Support

Principal Investigator:

Prof. Farah Emad Shamout, NYU Abu Dhabi (with co-PIs at NYU Grossman School of Medicine)

Project Overview

Objective:

Develop a multi-agent AI system that fuses electronic health records, medical imaging, and clinical notes to detect in-hospital patient deterioration earlier and more accurately than current single-modality approaches.

Key Innovations

  • Multimodal agentic architecture in which specialized AI agents reason jointly over EHR time-series, radiology imaging, and free-text clinical notes.

  • Hospital-grade evaluation and deployment pathway, co-designed with clinical partners to ensure workflow fit and safety.

  • Open tooling and reusable components that advance ADIA Lab’s broader multimodal AI research agenda.

Expected Outcomes

  • Earlier and more accurate identification of deteriorating inpatients.

  • Improved patient outcomes alongside more efficient staffing and resource allocation.

  • Open-source releases and hospital partnerships that extend ADIA Lab’s multimodal AI capabilities.

2. From GWAS to FWAS: Foundation Model–Wide Association Study of Clinical and Multi-Omics Phenotypes

Principal Investigator:

Prof. Eran Segal, Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), with co-Pis

Project Overview

Objective:

Move beyond traditional Genome-Wide Association Studies (GWAS) by leveraging pretrained DNA, RNA, and protein foundation models to systematically identify mechanistic gene–disease relationships across clinical and multi-omics phenotypes.

Key Innovations

  • A novel “Foundation model–Wide Association Study” (FWAS) framework that applies large-scale biological foundation models to phenotype discovery.

  • Integration of DNA, RNA, and protein representations to surface mechanistic, rather than purely statistical, signals.

  • Scalable pipelines designed to operate on diverse clinical and multi-omics datasets.

Expected Outcomes

  • Accelerated target discovery and improved disease risk prediction.

  • More precise interventions and higher R&D yield for biomedical and pharmaceutical applications.

  • Direct alignment with ADIA Lab’s push to accelerate biomedical sciences and drug discovery.

3. Causal AI for the Generation of Real-World Evidence from Healthcare Data

Principal Investigator:

Dr. Sophia Rein, Harvard University (with co-PIs)

Project Overview

Objective:

Build an open-source Causal AI toolkit that fuses the g-formula with modern deep learning to convert messy, longitudinal healthcare data into statistically valid cause-and-effect estimates at scale.

Key Innovations

  • Principled fusion of established causal inference methods (g-formula) with deep learning for high-dimensional, time-varying confounders.

  • An open-source software stack designed for reproducibility and adoption by health systems and regulators.

  • Cross-jurisdictional application using both U.S. and UAE healthcare data to test transportability of findings.

Expected Outcomes

  • Head-to-head, real-world comparative effectiveness analyses of GLP-1 receptor agonists versus alternative therapies on cardiovascular outcomes.

  • A reusable Causal AI toolkit for the broader real-world evidence community.

  • Direct support for ADIA Lab’s priorities on improving prevention and management of cardiometabolic diseases

More Information

For background on the 2025 call for research proposals, including the four thrust areas and evaluation criteria, please review the ADIA Lab Health Science RFP 2025.

Questions or collaboration opportunities? Reach out at healthsciences@adialab.ae.