A tool for translating polygenic scores onto the absolute scale using summary statistics.

Area of Science:

• Genetics and Bioinformatics
• Computational Biology
• Precision Medicine

Background:

• Polygenic scores (PS) show increasing predictive utility for health phenotypes.
• Safe clinical interpretation requires PS predictions on an absolute scale.
• Current methods often lack accessibility or require individual-level data.

Purpose of the Study:

• To develop and validate a method for converting polygenic scores to absolute scales for binary and continuous phenotypes.
• To evaluate methods for estimating polygenic score predictive utility (AUC/R^2) from genome-wide association study (GWAS) summary statistics.
• To provide tools for accurate and accessible interpretation of polygenic scores.

Main Methods:

• Developed a method using normal distribution theory to convert PS to absolute scales.
• Required only summary statistics (AUC/R^2, phenotype prevalence/distribution).
• Evaluated AUC/R^2 estimation from GWAS summary statistics using lassosum pseudovalidation.

Main Results:

• The absolute risk conversion method showed high concordance with observed values when AUC/R^2 was known.
• Lassosum pseudovalidation provided the most similar AUC/R^2 estimates to observed values.
• Deviations in AUC/R^2 estimates were noted for autoimmune disorders.

Conclusions:

• The developed method enables accurate interpretation of polygenic scores using only summary statistics.
• This facilitates educational and clinical applications of polygenic scores.
• Further research is needed to address barriers for clinical implementation, including GWAS sample representation.