tooluniverse-polygenic-risk-score

Query GWAS Catalog for genome-wide significant associations (p < 5e-8) for your trait, extract beta coefficients or convert odds ratios to log-odds, then compute a weighted sum of effect allele dosages. Production PRS should also apply LD clumping to avoid double-counting correlated variants.

Export your raw genotype file, map each rsID to its allele pair, then count effect allele dosages (0, 1, or 2) for each SNP in the PRS model. Multiply dosages by effect sizes and sum across SNPs, then standardize against population mean and standard deviation.

Yes, PRS accuracy drops substantially when GWAS discovery ancestry differs from the target population. A European-derived PRS can lose 50-70% of predictive power in West African populations due to differences in allele frequencies and linkage disequilibrium patterns.

PRS percentile indicates relative position in the population genetic risk distribution, not absolute disease probability. A high percentile means elevated risk compared to average, but most high-PRS individuals never develop the disease because environmental factors contribute substantially.

Use validated PGS Catalog models for clinical-grade applications because they include LD clumping, ancestry-specific weights, and external validation. Building a PRS from raw GWAS associations is appropriate for research exploration and educational purposes.

PRS captures only common variant effects, typically explaining 10-20% of trait variance while total heritability may be 50% or higher. Rare variants, epistasis, and gene-environment interactions are not captured by standard PRS models.