Extended Bayesian model averaging in generalized linear mixed models applied to schizophrenia family data.

Area of Science:

• Psychiatric Genetics
• Statistical Genomics

Background:

• Schizophrenia etiology research faces challenges due to complex interactions and diverse populations.
• Previous studies using generalized linear mixed models (GLMMs) have explored genetic and environmental influences on schizophrenia.
• Inconsistent findings highlight the need for robust methods to analyze multi-dimensional factors.

Purpose of the Study:

• To develop and validate an extended Bayesian model averaging (EBMA) procedure for schizophrenia.
• To estimate gene-gene (GG) and gene-environment (GE) interactions and heritability.
• To account for uncertainties in covariates and genetic model structures.

Main Methods:

• Employed an extended Bayesian model averaging (EBMA) procedure within GLMMs.
• Estimated variance components for random effects to assess heritability.
• Incorporated environmental and genetic covariates, and GG and GE interactions into competing models.
• Utilized simulation studies to compare EBMA with permutation tests and GEE methods.
• Applied the approach to data from singleton and multiplex schizophrenia families.

Main Results:

• EBMA demonstrated flexibility and stability in identifying candidate genes.
• The method successfully detected significant GE and GG interactions.
• Analysis adjusted for explanatory variables and intra-familial correlation structures.

Conclusions:

• EBMA is a valuable tool for dissecting complex genetic and environmental contributions to schizophrenia.
• The approach enhances the identification of susceptible genes and their interactions.
• This method offers a robust framework for analyzing familial aggregation in complex diseases.