General Framework for Meta-Analysis of Haplotype Association Tests

Area of Science:

• Genetics
• Statistical Genetics
• Genomic Epidemiology

Background:

• Complex traits are influenced by numerous genetic variants, often with small individual effects, necessitating large sample sizes for detection.
• Patient confidentiality limits direct data pooling, making meta-analysis crucial for combining results from multiple genetic studies.
• Existing meta-analysis methods primarily focus on single nucleotide variants (SNVs), with challenges in analyzing haplotypes formed by multiple SNVs across diverse studies.

Purpose of the Study:

• To develop and validate a robust two-stage meta-analysis approach for combining haplotype association results across multiple cohorts.
• To improve statistical power for detecting associations between haplotypes and complex traits, especially when considering low-frequency and rare genetic variations.
• To provide a framework for haplotype-specific and global association tests within a meta-analysis context.

Main Methods:

• A two-stage meta-analysis strategy was implemented: individual cohort haplotype effect estimation followed by multivariate generalized least square meta-analysis.
• The first stage involved regression-based estimation of haplotype effect sizes, incorporating relatedness adjustments where necessary.
• The second stage combined cohort-specific estimates using a multivariate generalized least square approach to yield overall haplotype effects and association tests.

Main Results:

• Simulation studies confirmed the proposed method controls type-I error rates effectively.
• The two-stage meta-analysis demonstrated superior power compared to inverse variance weighted meta-analysis of single SNV analyses when haplotype effects were present.
• A known association between the G6PC2 locus and fasting glucose was successfully replicated and refined using data from seven cohorts.

Conclusions:

• The proposed two-stage meta-analysis provides a powerful and reliable method for combining haplotype association studies.
• This approach enhances the ability to detect genetic associations for complex traits by leveraging multi-variant information.
• The method offers more precise effect estimates and facilitates robust haplotype-based genetic discoveries in large-scale consortia.