Fig. 2

Power and false-positive evaluation of the GBAT approach. a Power comparison of GBAT with a gene-based method Luijk et al., the top cis-eQTL approach where only the top cis-eQTL of a gene is tested for trans-eQTL, and the traditional SNP-based trans-eQTL scan. The cis-heritability was set to 0.1, and the per-gene trans-heritability was set to 0.02. Power was assessed at 5% FDR using BH correction and was computed as the fraction of 2000 simulations. Colors represent different methods. b Quantile-quantile plot of trans-association p values from permutation analyses of GBAT, Luijk, and top cis-eQTL methods. The cis-heritability was set to 0.1, the trans-heritability was set to 0.02, sample size is 900, and the causal proportion is at 1%. The simulated expression of the trans-gene is randomly permuted for each simulation. The plot is based on 2000 simulations