Fig. 2

Model selection. Models for estimating the PGS were constructed employing either the linkage disequilibrium pruning and thresholding method (A) or LDpred [21] (B), using the effect sizes of variants in a publicly available genome-wide association study summary for BMI [20]. Individual PGS was ascertained using candidate models constructed within the TMM10K dataset. The AUC for obesity (BMI ≥ 25) was assessed in R package “ROCR,” with age and sex as covariates. The LDpred model at ρ = 0.03 exhibited the highest AUC and was used in subsequent analyses. PGS polygenic score, AUC area under the curve, BMI body mass index, TMM Tohoku Medical Megabank