Fig. 1: Simulation study for the performance of our BayesRR-RC model implemented in the GMRM software against existing approaches for variance component and genomic annotation enrichment estimation. | Nature Communications

Fig. 1: Simulation study for the performance of our BayesRR-RC model implemented in the GMRM software against existing approaches for variance component and genomic annotation enrichment estimation.

From: Probabilistic inference of the genetic architecture underlying functional enrichment of complex traits

Fig. 1

a Correlation of the simulated and estimated SNP heritability across 13 genomic annotation groups within each of 20 replicates for five different statistical models: a mixture of regression model with multiple group-specific variance components described in this work (GMRM), Haseman–Elston regression with annotation-specific relationship matrices implemented in the RHEmc software (RHEmc), a multiple group-specific variance component REML model implemented in the software bolt (BOLT), and two annotation summary statistic models implemented in the software LDSC and LDAK. The column facets give the simulated heritability and rows give the number of causal variants. b Probability of falsely assigning one of the 13 genomic annotation groups as explaining 2 times greater proportion of variance given the proportion of SNPs mapping to the annotation. The column facets give the simulated heritability and rows give the number of causal variants. Boxplots give the median with 25th and 75th percentile and 95% credible intervals for n = 20 simulation replicates in both panels.

Back to article page