Contemporary Considerations for Constructing a Genetic Risk Score: An Empirical Approach

Area of Science:

• Genetics
• Biostatistics
• Cardiovascular Disease Research

Background:

• Genetic risk scores (GRS) aggregate Single Nucleotide Polymorphisms (SNPs) for disease risk prediction.
• Current GRS often use only genome-wide significant SNPs, potentially missing valuable information.

Purpose of the Study:

• To determine the optimal set of SNPs for constructing a robust genetic risk score.
• To evaluate the impact of various parameters on GRS performance using the ARIC Study cohort.

Main Methods:

• Investigated P-value thresholds, linkage disequilibrium, imputation quality, and imputation type for SNP selection.
• Utilized diverse evaluation metrics to assess GRS performance for coronary heart disease.

Main Results:

• The P-value threshold demonstrated the most significant impact on GRS quality, with an optimal value near 0.001.
• GRSs exhibited robustness to variations in linkage disequilibrium and imputation quality.
• Optimal GRS construction is influenced by the chosen evaluation metric and intended application.

Conclusions:

• Empirically selecting SNPs enhances GRS robustness and predictive capability.
• The optimal GRS strategy is adaptable based on specific research goals and evaluation criteria.