Figure 1: ME-Bayes SL Workflow.

[Step 0] apply LDpred2 to each of the K training populations (ancestry groups) to obtain estimated causal SNP proportions $\left(p_k,k=1,\dots ,K\right)$ and heritability $\left(h_k^2,k=1,\dots ,K\right)$ parameters based on the tuning set, these parameters will be used to specify the prior distributions and tuning parameter settings for ME-Bayes. [Step 1] ME-Bayes: jointly model across all training populations to obtain a total of $(\mathrm{L}\times \mathrm{K})$ PRS models under $\mathrm{L}$ different tuning parameter settings for $\mathrm{P}\mathrm{r}\left({\delta }_{1j},\dots ,{\delta }_{Kj}\right)$ (functions of $p_k\mathrm{s}$) and ${\rho }_{k_1{k}_2}\mathrm{s}$ across $\mathrm{K}$ training populations. [Step 2] for the target population, apply the super learning (SL) algorithm with 3 base learners (elastic net regression, ridge regression, and linear regression) to train an “optimal” linear combination of the $(\mathrm{L}\times \mathrm{K})$ PRS models, which we call the ME-Bayes SL PRS model, based on the tuning set of the target population. The prediction performance of the final ME-Bayes SL PRS model should be evaluated on an independent testing set.