Fig. 2: Comparison of performance in simulations using the CHURC1 gene as an example. | Nature Communications

Fig. 2: Comparison of performance in simulations using the CHURC1 gene as an example.

From: SUMMIT: An integrative approach for better transcriptomic data imputation improves causal gene identification

Fig. 2

Plots of imputation R2 (a) and subsequent power (b) in test samples by SUMMIT, Lassosum, TWAS-fusion, and PrediXcan, with varying expression heritability \({h}_{e}^{2}\) and proportion of true causal SNPs pcausal. The results were based on 1000 simulation replicates. In subfigure (a), the box limits represent the lower and upper quartiles, the central line represents the median, and the whiskers represent all samples lying within 1.5 times the interquartile range (IQR). Subfigure (c) shows the relationship between the expression panel sample size and TWAS power or expression prediction accuracy using SUMMIT. For subfigures (b) and (c), we set \({h}_{p}^{2}=0.2\); for subfigure (c), we set pcausal = 0.05. p-values were calculated by the two-sided tests; empirical power was estimated by the proportions of p-values less than the significance threshold 2.5 × 10−6. The empirical power comparisons for \({h}_{p}^{2}\in (0.1,0.5,0.8)\) are shown in Supplementary Fig. 1.

Back to article page