Table 5:
Assessment of concordance for each preprocessing method based on the compendium of 30 experiments
|
t2-statistic |
Bayesian network |
|||||
|---|---|---|---|---|---|---|
| FDR | 0.01 | 0.05 | 0.25 | 0.01 | 0.05 | 0.25 |
| RMA—default | 2.73 | 2.60 | 2.57 | 3.40 | 3.00 | 2.47 |
| RMA—no BG | 3.13 | 2.80 | 2.53 | 3.50 | 2.83 | 2.33 |
| PLIER—no MM | 3.50 | 3.23 | 3.03 | 3.47 | 2.70 | 2.80 |
| fRMA—default | 4.53 | 4.23 | 4.03 | 4.27 | 4.03 | 3.93 |
| MBEI—default | 5.47 | 5.70 | 5.73 | 5.37 | 5.47 | 6.00 |
| MBEI—no MM | 5.90 | 5.93 | 6.37 | 5.43 | 5.73 | 5.77 |
| GCRMA—default | 6.47 | 6.80 | 6.87 | 6.03 | 6.33 | 6.13 |
| PLIER—default | 5.87 | 6.10 | 6.27 | 6.90 | 7.67 | 8.30 |
| MAS5.0—default | 7.40 | 7.60 | 7.60 | 6.63 | 7.23 | 7.27 |
| CDF of highest ranked method | ||||||
| P(rank ≤ 1) | 0.96 | 0.76 | 0.50 | 0.49 | 0.60 | 0.66 |
| P(rank ≤ 2) | 1.00 | 1.00 | 0.95 | 0.77 | 0.85 | 0.97 |
| P(rank ≤ 3) | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
The upper portion of the table shows the average rank of the concordance scores across the compendium of data sets. A lower rank corresponds to a larger number of gene sets called differentially expressed using other preprocessing methods also being called differentially expressed using the given preprocessing method. The lower portion of the table displays the CDF of the highest ranked method, providing an assessment of our confidence in the highest ranking. Note: MM, mismatch; BG, background correction.