Evaluating ancestry adjustment in multi-ancestry epigenome-wide analysis.

Area of Science:

• Genetics
• Epigenetics
• Population Genetics

Background:

• Accurate adjustment for population substructure is crucial in epigenome-wide association studies (EWAS), especially in diverse ancestral cohorts.
• The EPISTRUCTURE tool, a genotype-free ancestry inference method, was initially validated using a European reference population (KORA).
• Its performance in multi-ancestry settings requires further evaluation.

Purpose of the Study:

• To assess the effectiveness of EPISTRUCTURE's principal components (PCs) for ancestry adjustment in EWAS using multi-ancestry cord-blood samples.
• To compare ancestry adjustment using a standard European (KORA) reference versus novel references derived from the Hyperglycemia and Adverse Pregnancy Outcome (HAPO) cohort.
• To evaluate the impact of different reference sets on association detection and biological interpretation for newborn adiposity traits.

Main Methods:

• Systematically evaluated EPISTRUCTURE PCs derived from KORA and HAPO-based reference sets.
• HAPO reference sets were constructed using varying SNP-CpG R² thresholds (e.g., RS30: R² > 0.3) to identify ancestry-informative CpGs.
• Applied these references for population substructure adjustment in EWAS of birthweight, cord C-peptide, and sum of skinfolds in HAPO cord blood samples (n=3,116).

Main Results:

• The HAPO RS30 reference consistently yielded lower genomic inflation compared to the KORA reference.
• Biologically relevant associations for birthweight and cord blood C-peptide were more frequently identified using the HAPO-derived reference.
• Pathway enrichment analyses revealed significant immune and metabolic signals, with some pathways uniquely identified using the HAPO reference.

Conclusions:

• Reference sets generated from multi-ancestry cohorts, such as HAPO, provide superior ancestry adjustment in EWAS.
• Utilizing ancestry-informative CpGs from diverse populations enhances the detection of true biological associations.
• This approach improves the reliability and interpretability of EWAS findings in genetically heterogeneous cohorts.