Meta-analysis of Complex Diseases at Gene Level with Generalized Functional Linear Models

Area of Science:

• Statistical Genetics
• Genomic Association Studies
• Bioinformatics

Background:

• Meta-analysis of genetic data is crucial for identifying associations between genetic variants and complex traits.
• Existing methods like MetaSKATs utilize mixed-effect models, which may not be optimal for all genetic architectures.
• There is a need for flexible statistical models that can handle various genetic variant frequencies (rare and common) in meta-analyses.

Purpose of the Study:

• To develop and evaluate Generalized Functional Linear Models (GFLMs) for meta-analysis of multiple case-control studies.
• To assess the performance of GFLMs in detecting associations between genetic data and dichotomous traits, adjusting for covariates.
• To compare the statistical power and type I error rates of GFLM-based tests against existing methods like MetaSKATs.

Main Methods:

• Developed Generalized Functional Linear Models (GFLMs) as fixed models for genetic meta-analysis.
• Derived chi-squared-distributed Rao's efficient score test and likelihood-ratio test (LRT) statistics based on GFLMs.
• Conducted extensive simulations to evaluate type I error rates and statistical power.
• Applied GFLMs to a meta-analysis of type 2 diabetes genetic data from eight European studies.

Main Results:

• GFLM-based Rao's efficient score tests showed conservative type I error rates and higher power than MetaSKATs when causal variants were a mix of rare and common.
• LRT statistics provided accurate type I error rates for homogeneous genetic-effect models but could inflate rates for heterogeneous models.
• Application to type 2 diabetes data identified significant associations for 18 gene regions, outperforming MetaSKATs which detected none.
• GFLMs successfully analyzed a combination of rare and common variants.

Conclusions:

• GFLMs provide a robust and powerful framework for genetic meta-analysis of dichotomous traits, accommodating diverse genetic variant frequencies.
• The developed Rao's efficient score test statistics offer advantages in power over MetaSKATs under specific genetic architectures.
• GFLMs are valuable tools for large-scale genetic studies, including whole-genome and whole-exome association studies, enhancing the detection of trait-associated genes.