Type 2 diabetes (T2D) is a polygenic metabolic disorder for which GWAS have identified over 600 risk loci, but the causal genes and the tissues mediating those associations remain poorly resolved. This study applied tissue-specific, ancestry-aware transcriptome-wide association studies (TWAS) to refine gene-level signals at T2D GWAS loci and to evaluate tissue context and cross-ancestry replication.
The authors performed TWAS in six tissues relevant to T2D biology: subcutaneous adipose, visceral adipose, brain hypothalamus, liver, skeletal muscle, and pancreas. For European-ancestry (EUR) analyses, summary statistics came from the largest available EUR T2D GWAS (242,283 cases and 1,569,734 controls). Expression prediction models were pre-trained using GTEx data from 689 EUR individuals.
An African American (AFA) TWAS was also conducted to assess cross-ancestry replication. The AFA analysis used summary statistics from the largest AFA T2D GWAS (50,251 cases and 103,909 controls) and gene expression models trained in 111 AFA individuals in GTEx.
All TWAS analyses included downstream conditional analyses to identify independent gene-level signals at loci. Significance thresholds and the exact statistical procedures used to derive P values were reported in the study (significance threshold P < 1.919 x 10-6 reported for gene–T2D associations).
Across the six tissues, the EUR-TWAS identified between 684 and 750 significant gene–T2D associations per tissue at the reported significance threshold. These associations implicated both established T2D candidate genes and novel candidates. The results highlight the capacity of tissue-specific TWAS to prioritize gene targets at loci discovered by GWAS and to suggest which tissues may mediate genetic risk.
The authors used conditional analyses to refine signals and remove correlated associations arising from linkage disequilibrium and co-regulation. After conditioning, each tissue retained between 289 and 322 independent TWAS signals, indicating substantial regulatory heterogeneity and multiple independent gene-level effects within T2D loci.
To evaluate robustness across ancestries, the study tested whether EUR-derived TWAS signals were enriched in the AFA-TWAS. Significant enrichment was observed for subcutaneous adipose, visceral adipose, skeletal muscle, and pancreas, supporting cross-ancestry replication of many EUR-derived associations. Enrichment in liver was weaker, a result the authors attribute to the limited sample size of the AFA GTEx expression models and the AFA GWAS data relative to EUR resources.
Some genes showed consistent association across all six tissues; notably JAZF1 and IDE were associated with T2D in every tissue analyzed. Other genes exhibited tissue-restricted patterns: TCF7L2 and WSF1 demonstrated heterogeneous effects limited to a subset of the examined tissues. These examples illustrate how integrating tissue context can distinguish broadly acting regulatory genes from tissue-specific contributors to disease risk.
The authors explicitly highlight limitations arising from the scarcity of non-European transcriptomic datasets. GTEx-derived expression models for non-EUR ancestries were smaller (111 AFA donors) than EUR models (689 donors), which likely reduced power for replication in some tissues (for example, liver). The study calls for larger, more diverse ancestry-specific transcriptomic resources to improve ancestry-aware TWAS and to ensure that signals are robust across populations.
All data produced in the study are available upon reasonable request to the authors. The analyses relied on GTEx data and the manuscript reports that necessary ethical approvals and participant consents were obtained through GTEx. The authors declared no competing interests.
Declared funders include the Medical Research Council (grant MR/V020749/1) and the National Institute for Health and Care Research (NIHR) Manchester Biomedical Research Centre (NIHR203308). The authors reported no competing interests.
Integrating tissue-specific and ancestry-aware TWAS can refine gene assignments at T2D GWAS loci, reveal tissue-context effects, and identify both shared and tissue-restricted candidate genes. Cross-ancestry replication in African American cohorts supports the robustness of many EUR-derived signals in adipose, muscle, and pancreas, but weaker replication in liver emphasizes the impact of limited non-EUR sample sizes. The authors recommend expanded, diverse transcriptomic resources to improve causal gene discovery and to generalize findings across ancestries.
Note: This report summarizes findings presented in the cited preprint. The preprint has not been peer reviewed and should not be used to guide clinical practice.