New AGEMdb Database Expands Ancestry-Specific Gene Expression Models for Genomic Studies

A new web-based database named AGEMdb has been introduced to provide ancestry-specific cross-tissue gene expression imputation models, addressing a critical lack of diversity in genomic research resources. Published in BMC Bioinformatics by researchers Inti Pagnuco, Francisco Herrerías-Azcué, Stephen Eyre, Magnus Rattray, and Andrew P. Morris, the open-access platform delivers models trained specifically for underrepresented populations. By making expression imputation models tailored to African American individuals freely available alongside models for European populations, the developers aim to support more equitable gene–trait association analyses and promote inclusive transcriptome-wide studies.

The development of AGEMdb targets an established methodological need within transcriptome-wide association studies, known as TWAS. In genomic research, TWAS integrate genetically regulated gene expression with genome-wide association study (GWAS) data to identify gene–trait associations. These analyses rely on accurate gene expression prediction models, which must be trained using genotype and transcriptomic data collected from the same set of individuals. Consequently, the reliability and utility of TWAS investigations depend fundamentally on the quality and composition of the datasets from which these predictive models are derived.

Currently, the vast majority of publicly available TWAS models are based on reference resources that predominantly consist of participants of European ancestry. According to the authors, this heavy European ancestry bias substantially limits the transferability and generalizability of existing gene expression models across diverse populations. Because existing predictive tools reflect this demographic skew, applying European-trained models to diverse ancestral cohorts can undermine the accuracy and equity of subsequent gene–trait analyses.

To resolve this imbalance, the research team analyzed genotype and gene expression data obtained from the Genotype-Tissue Expression (GTEx) Project. Following rigorous quality control and confounder adjustment, the investigators curated distinct datasets comprising 689 European ancestry individuals and 111 African American individuals. Using these curated GTEx datasets, the researchers then developed ancestry-specific cross-tissue gene expression imputation models using the UTMOST framework, establishing separate computational models for both ancestral groups across tissues.

The resulting models are now accessible through AGEMdb, an interactive online database that allows researchers to browse, search, and download gene expression imputation models with flexible filtering options, such as filtering by tissue and gene. By incorporating African American-specific models into public research workflows, AGEMdb directly addresses the European ancestry bias found in earlier TWAS resources. The authors conclude that the platform provides a valuable resource for ancestry-specific TWAS, promoting more inclusive studies that support the advancement of equitable genomic research.

Subject of Research: Ancestry-specific gene expression imputation database for transcriptome-wide association studies

Article Title: Database for ancestry-specific cross-tissue gene expression models: AGEMdb

Article References: Pagnuco, I., Herrerías-Azcué, F., Eyre, S., Rattray, M., & Morris, A. P. (2026). Database for ancestry-specific cross-tissue gene expression models: AGEMdb. BMC Bioinformatics. https://doi.org/10.1186/s12859-026-06743-2

Image Credits: AI Generated

DOI: 10.1186/s12859-026-06743-2

Keywords: AGEMdb, transcriptome-wide association studies, gene expression models, GTEx Project, UTMOST framework, genomic diversity, bioinformatics, computational genetics, Database, ancestry-specific, cross-tissue, gene