File:LDpred2 example.png

English: (a) A comparative representation of the key parameters (p, h2, and α) utilized in the LDpred2 and the enhanced LDpred2-auto Bayesian models. The original LDpred2, which does not account for the α parameter (indicating its value as 0), is contrasted against the LDpred2-auto, which integrates α as a new parameter to model negative selection. (b) Inference of out-of-sample predictive performance illustrated using two sets of effect sizes (β1 and β2) derived from independent chains of the Gibbs sampler in LDpred2-auto. The gradient color spectrum indicates the deviation magnitude between β1 and β2, emphasizing areas where the effect sizes closely mirror each other (minimal deviation) versus areas of substantial difference. The close alignment between effect sizes would imply good prediction accuracy.