A strategy analysis for genetic association studies with known inbreeding

Area of Science:

• Genetics
• Statistical genomics
• Population genetics

Background:

• Genome-wide association studies (GWAS) are effective for Mendelian diseases but less so for complex diseases due to gene-gene and gene-environment interactions.
• Increasing genotype data (millions of markers, imputation, sequencing) introduces noise, necessitating larger sample sizes and advanced analytical methods.
• Current methods often overlook crucial gene-gene and gene-environment interactions, particularly in complex diseases.

Purpose of the Study:

• To propose a non-parametric additive model for detecting genetic variants associated with diseases, accounting for unknown order interactions.
• To enhance additive models by incorporating genealogical information in isolated populations, where relatedness correlates with genetic similarity.
• To identify disease-associated genes and environmental variables using Random Forest (RF) after selecting highly inbred cases and controls via the Hungarian method.

Main Methods:

• Utilized a non-parametric additive model to analyze genetic association.
• Incorporated genealogical data from isolated populations to improve model performance.
• Employed the Hungarian method for optimal selection of highly inbred cases and controls.
• Applied Random Forest (RF) for estimating disease-associated genetic and environmental variables.

Main Results:

• Developed a procedure to effectively eliminate stratification between cases and controls, enhancing precision in identifying disease-responsible genetic variants.
• Successfully applied the method to beta-thalassemia (Mendelian disease) and common asthma (complex disease).
• Identified candidate genes for common asthma susceptibility, some of which are supported by existing literature.

Conclusions:

• The data analysis approach, combining case-control selection with RF, is a powerful tool for detecting disease-associated genetic variants in isolated populations.
• The method yields a predictive model with high accuracy for estimating unknown disease status.
• This approach can be generalized for developing diagnostic test kits for various Mendelian diseases.