Connect: cohort-oriented network knowledgebase for enhanced collaboration

The National Institute of Environmental Health Sciences awarded $605,091 to the University of Texas at Arlington on April 16, 2026, under the Environmental Health grant program (CFDA 93.113) to develop CONNECT (Cohort-Oriented Network Knowledgebase for Enhanced Collaboration), a directed knowledge graph summarizing environmental health research findings from children's cohort studies.

The recipient will execute three primary aims: develop an AI-assisted knowledge extraction pipeline with prompt engineering to harmonize environmental health knowledge from published cohort studies; build the CONNECT knowledgebase for children's environmental health; and develop a web tool showcasing the knowledgebase's utility for users. The project addresses the need to make cohort study findings and knowledge easily available and accessible to the environmental health community and public, harmonizing existing cohort study knowledge in children's health while supporting future environmental health research directions for managing multi-cohort findings.

Performance runs through April 15, 2029, with work conducted in Arlington, Texas. The University of Texas at Arlington's Office of Research Administration serves as the prime recipient.

Sub-recipients deliver collaborative research support. Columbia University's Health Sciences Division received $103,899 in a sub-award executed May 18, 2026. The University of California, Los Angeles Office of Research Administration received $103,899 in a sub-award executed May 15, 2026. The source does not describe the specific contributions of the sub-recipients to the knowledgebase development.

Abstract

Cohort studies play a fundamental role in contemporary environmental health sciences, advancing our understanding of how chemical exposures, lifestyle factors, and genetics contribute to human health. The national institutes of health (NIH) has always recognized the importance of cohort studies, especially in enhancing collaboration by pulling cohort resources to include participants from various populations and backgrounds. Summarizing and managing the findings and knowledge from cohort studies are indispensable components of cohort development, yet these tasks have often been overlooked. IT will inevitably result in a significant loss if the knowledge and findings from cohort studies are not easily available or accessible to the entire environmental health community and public. Therefore, in this project, we propose cohort-oriented network knowledgebase for enhanced collaboration (connect), a directed knowledge graph summarizing the knowledge for children's environmental health, to meet the emerging needs of contemporary environmental health science community. The connect knowledgebase will harmonize existing cohort study knowledge in children's health and align with future environmental health development directions for managing multi-cohort findings. Indeed, the proposed connect knowledgebase is a unique resource that is distinguished from other existing knowledgebases in the environmental health community. Our aims are: AIM1. Develop an AI-assisted knowledge extraction pipeline with prompt engineering for harmonizing environmental health knowledge from published cohort studies; aim 2: develop the connect knowledgebase for children's environmental health; AIM3: develop connect knowledgebase webtool and showcase the utility of connect knowledgebase for users. Specifically, we will first develop an AI-assisted pipeline for knowledge extraction, validation, and harmonization from publications. We will leverage chatgpt to assist in extracting findings/associations from 17 niehs funded children's cohort publications. We will harmonize our environmental health language based on the existing hhear ontology. Secondly, we will further construct the connect knowledgebase based on the extracted knowledge using the proposed pipeline. The connect knowledgebase will consist of knowledge from each cohort, resulting in an overall knowledge graph with harmonized environmental health language. Moreover, we will develop a webtool to facilitate knowledge dissemination to the environmental health community. We also demonstrate the various utilities of the connect knowledgebase, including exposure-outcome evidence mapping, mechanistic insights and hypothesis generation, statistical querying for meta-analysis, among many other applications. Finally, our proposed connect knowledgebase will provide a unique opportunity to train undergraduate students to become the next generation biomedical scientists in both data science and environmental health.