Table 1.
Power (%) | FDR (%) | F1 score (%) | Power (strong) (%) | Power (weak) (%) | |
---|---|---|---|---|---|
EBSeqHMM | 98.6 | 4.3 | 97.1 | 99.7 | 97.5 |
EBSeq | 90.0 | 0.1 | 94.7 | 93.9 | 86.1 |
DESeq2 | 92.4 | 0 | 96.1 | 95.4 | 89.4 |
edgeR | 92.5 | 0.1 | 96.1 | 96.1 | 89.4 |
voom | 91.9 | 0 | 95.8 | 95.1 | 88.6 |
maSigPro (0.7) | 46.8 | 0 | 63.8 | 56.1 | 37.5 |
maSigPro (0.5) | 76.1 | 0.1 | 86.4 | 81.5 | 70.6 |
maSigPro (0.3) | 86.9 | 0.5 | 92.8 | 90.6 | 83.2 |
FC (2.5) | 0.6 | 0.2 | 1.2 | 0.8 | 0.5 |
FC (2) | 3.4 | 1.4 | 6.6 | 4.3 | 2.6 |
FC (1.5) | 42.1 | 3.5 | 58.7 | 55.7 | 28.6 |
FC (1.3) | 90.0 | 8.5 | 90.7 | 97.5 | 82.4 |
FC (1.2) | 98.6 | 19.7 | 88.6 | 99.8 | 97.9 |
The first three columns show the average power, FDR and F1 score for detecting DE genes in Sim I. Power within the strong and weak groups is further evaluated in columns 4 and 5. Averages are calculated over 100 Sim I simulations. The standard errors (not shown) for EBSeq-HMM, EBSeq, DESeq2, edgeR, voom and maSigPro (and in most cases FC) were .