Score tests for familial correlation in genotyped-proband designs

Area of Science:

• Genetics
• Epidemiology
• Statistical genetics

Background:

• The genotyped-proband design involves observing a proband's phenotype and genotype, then relatives' phenotypes, but not their genotypes.
• Existing analyses assume conditional independence of phenotypes given genotypes, an assumption that may not always hold.
• Developing tests for conditional independence is crucial for robust genetic analyses.

Purpose of the Study:

• To develop a score test for the conditional independence assumption in genetic studies.
• To address models with covariates or partial observation of relatives' genotypes.
• To provide a robust and simple analytical approach for missing covariate data.

Main Methods:

• The study frames the problem as score testing in the presence of missing covariates.
• A natural conditional likelihood is used to develop the score test.
• The proposed method avoids the need to specify covariate distributions, enhancing model-robustness.

Main Results:

• A simple score test is derived for conditional independence.
• The method is applicable to complex models including covariates and partially observed genotypes.
• The approach is demonstrated using data on breast cancer penetrance for BRCA1 and BRCA2 mutations.

Conclusions:

• The developed score test offers a robust and straightforward method for assessing conditional independence in genetic epidemiology.
• This approach is valuable for studies with missing genotype data, improving analytical accuracy.
• The methods are practically applied to understanding breast cancer risk associated with BRCA mutations in specific populations.