Large-Scale Genotype-Based Trait Imputation With Multi-Ancestry GWAS Data

Area of Science:

• Genetics
• Bioinformatics
• Population Genetics

Background:

• Genome-wide association studies (GWAS) identify genetic variants for complex traits but often lack diversity, limiting generalizability.
• LS-Imputation enhances GWAS by imputing traits from summary statistics but struggles with accuracy in underrepresented ancestries due to smaller sample sizes.
• Alzheimer's disease (AD) research requires diverse genetic data to understand its complex heritability across populations.

Purpose of the Study:

• To develop and evaluate novel LS-Imputation methods that integrate multi-ancestry GWAS data for improved trait imputation accuracy.
• To enhance the performance of trait imputation in non-European populations, addressing limitations of existing methods.
• To facilitate genetic association analyses for complex diseases like AD in diverse ancestral groups.

Main Methods:

• Proposed two novel LS-Imputation variants: LS-Imputation-Combined and LS-Imputation-Transfer.
• LS-Imputation-Combined merges GWAS summary statistics from multiple ancestries.
• LS-Imputation-Transfer uses stochastic gradient descent for sequential imputation refinement across ancestries.

Main Results:

• Integrating multi-ancestry GWAS data significantly improved trait imputation accuracy compared to single-ancestry approaches.
• LS-Imputation-Transfer demonstrated the highest imputation performance across evaluated datasets.
• Proof-of-concept using high-density lipoprotein (HDL) cholesterol levels was successful before application to AD status imputation.

Conclusions:

• Novel LS-Imputation methods effectively leverage multi-ancestry GWAS data to boost imputation accuracy.
• LS-Imputation-Transfer shows particular promise for improving genetic studies in diverse populations.
• Enhanced imputation accuracy supports more robust genetic association analyses for complex diseases like Alzheimer's disease across ancestries.