Analysis of genome-wide association studies with multiple outcomes using penalization

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Genome-wide association studies (GWAS) aim to identify genetic markers for complex traits.
• Penalization methods are used for marker selection with numerous single nucleotide polymorphisms (SNPs).
• Existing methods often analyze single phenotypes, failing to account for correlations among multiple related outcomes.

Purpose of the Study:

• To develop a statistical method for joint analysis of multiple correlated phenotypes in GWAS.
• To identify shared genetic markers associated with all outcome variables.
• To address limitations of single-response penalization methods in complex genetic datasets.

Main Methods:

• Employed joint modeling to account for correlations among multiple response variables.
• Utilized the group Lasso approach for selecting markers associated with all phenotypes.
• Developed an efficient computational algorithm for the proposed method.
• Validated the approach using simulation studies and a heterogeneous stock mice dataset.

Main Results:

• The proposed group Lasso method effectively identified shared significant SNPs across multiple correlated phenotypes.
• The method demonstrated superior performance compared to existing penalization techniques in simulation studies.
• Analysis of the heterogeneous stock mice dataset revealed biologically relevant marker associations.

Conclusions:

• The developed joint modeling and group Lasso approach is effective for GWAS with multiple correlated phenotypes.
• This method offers an advancement over single-response models for complex trait genetics.
• The approach provides a robust tool for genetic marker discovery in large-scale studies.