Fig. 4: Simulation results of various multi-population PRS methods in large training sample sizes and large causal SNP proportions. | Nature Communications

Fig. 4: Simulation results of various multi-population PRS methods in large training sample sizes and large causal SNP proportions.

From: JointPRS: A data-adaptive framework for multi-population genetic risk prediction incorporating genetic correlation

Fig. 4: Simulation results of various multi-population PRS methods in large training sample sizes and large causal SNP proportions.

a–d, The total heritability was fixed at h2 = 0.4. The proportion of causal SNPs was set to p = 0.1 and 0.01. We evaluated four non-European populations, (a) EAS, (b) AFR (c) SAS, (d) AMR. The training sample sizes for non-European populations was set to Ntrain = 80, 000. The tuning sample sizes for the target population was set to Ntune = 500, 2000, 5000, and 10,000. Simulations in each scenario were repeated five times, and the mean values was presented in the bar plots. All seven methods were considered here, with the best and second-best methods denoted by two stars and one star, respectively, in the corresponding bar plots. Note that SDPRX cannot provide predictions for SAS and AMR as it does not provide the corresponding LD reference panels. Source data are provided as a Source Data file.

Back to article page