Table 1.
Highlighting issues with not correcting for phylogeny
| Method | Correlation | TPR (S.D) | FPR (S.D) | FDR (S.D) | Overall Accuracy (S.D) |
|---|---|---|---|---|---|
| multivariate BM PCM (p<0.05) | ρC | 0.476 (0.0004) | 0.026 (0.0030) | 0.053 (0.0056) | 0.725 (0.0013) |
| cor.test() (p<0.05) | ρU | 0.574 (0.0006) | 0.209 (0.0075) | 0.267 (0.0068) | 0.682 (0.0035) |
| Fraser et. al. [8] | ρC | 0.567 (0.0363) | 0.053 (0.0152) | 0.084 (0.0212) | 0.757 (0.0148) |
| ρU | 0.511 (0.0432) | 0.097 (0.0285) | 0.156 (0.0316) | 0.708 (0.0144) | |
| Martin and Fraser [12] | ρC | 0.476 (0.0108) | 0.025 (0.0008) | 0.050 (0.0010) | 0.726 (0.0051) |
| ρU | 0.305 (0.0155) | 0.016 (0.0010) | 0.050 (0.0015) | 0.644 (0.0073) |
Comparison of 4 methods for detecting coevolution of gene expression using data simulated under Brownian Motion. The 4 methods represent the multivariate Brownian Motion (BM) PCM described in this manuscript, hypothesis testing with the phylogenetically-uncorrected correlation, the method described in Fraser et. al. [8], and the method described in Martin and Fraser [12]. Mean and standard deviations for true positive rates (TPR), false positive rates (FPR), false discovery rate (FDR), and overall accuracy are reported