SummaryAUC

SummaryAUC estimates the area under the receiver operating characteristic curve (AUC) and its variance for polygenic risk scores using genome-wide association study (GWAS) summary statistics from validation datasets.

Key Features:

• Approximation of AUC: Approximates AUC for polygenic risk scores (PRS) using summary-level GWAS statistics from validation datasets.
• Variance Estimation: Estimates the variance of the AUC to quantify uncertainty in PRS performance.
• Compatibility with genotyped or imputed SNPs: Applicable to PRSs composed of single nucleotide polymorphisms (SNPs) that are either genotyped or imputed in the validation dataset.
• Computational efficiency: Reduces computational demands compared to approaches requiring individual-level genotype and phenotype data.
• Empirical validation: Demonstrated high accuracy with minimal bias (typically <0.5%) in tests using large-scale GWAS data.
• Genome-wide SNP limitation: Not applicable to PRSs that include all genome-wide SNPs due to computational constraints.

Scientific Applications:

• Complex disease risk prediction: Evaluation of PRS performance for complex diseases, including schizophrenia, using summary statistics.
• PRS development and validation: Facilitates AUC approximation during development and validation of PRSs when individual-level data are unavailable.
• GWAS-based validation: Enables validation of risk prediction models directly from GWAS summary statistics from validation cohorts.

Methodology:

Statistical methods leverage GWAS summary statistics from validation datasets to approximate AUC and estimate its variance; methods were evaluated on large-scale GWAS data showing high accuracy with minimal bias (typically <0.5%).