Table 2.
Summary of quality measures for temperature calibration.
| Hyb. Temp., k | accfly | Ifly |
|
nsig | accspk | Ispk |
|
||
|---|---|---|---|---|---|---|---|---|---|
| Hyb 47 C | 71.74 | 0.405 | -67616 | 2925 | 71.05 | 0.106 | -19780 | ||
| Hyb 49 C | 72.25 | 0.417 | -65697 | 1979 | 74.14 | 0.234 | -16810 | ||
| Hyb 50 C (a) | 74.06 | 0.451 | -60972 | 3201 | 71.94 | 0.133 | -18930 | ||
| Hyb 50 C (b) | 75.05 | 0.461 | -60111 | 3628 | 73.28 | 0.261 | -16454 | ||
| Hyb 51 C | 75.06 | 0.467 | -58616 | 3810 | 74.10 | 0.341 | -14683 | ||
| Hyb 52 C | 71.79 | 0.402 | -68745 | 2140 | 73.16 | 0.156 | -18357 | ||
| Hyb 54 C | 70.35 | 0.385 | -70643 | 2348 | 68.34 | 0.184 | -18371 | ||
| Hyb 56 C | 64.25 | 0.292 | -83426 | 1213 | 60.64 | 0.094 | -20323 | ||
The left part of the table shows results from the protocol assessment using biological samples for calibration ('fly'). In comparison, the corresponding results for spiked-in exogenic RNAs are shown on the right ('spk'). Column 'Hyb. Temp., K' lists the considered protocol K and its hybridization temperature. For each protocol, the table displays the achieved generalization accuracies ('acc'), the average mutual information I, the log likelihoods log 
and the number n of genes with differential expression calls by FSPMA ANOVA at a 1% FDR threshold. The generalization accuracy reflects protocol performance for equal misclassification costs. ROC curves are provided for the generic case (Fig. 4). Detailed results for each protocol can be found in the Supplement.