Table 1 |
|||||||||
|
False discovery rates for present and marginally present classifications |
|||||||||
| 5% |
10% |
17% |
|||||||
|
|
|
|
|||||||
| P value |
FP |
TP |
P value |
FP |
TP |
P value |
FP |
TP |
|
|
|
|||||||||
| Raw PM and MM |
0.008 |
1,098 |
19,398 |
0.025 |
2,291 |
20,182 |
0.060 |
4,300 |
20,687 |
| GC-RMA PM and MM |
0.024 |
1,127 |
20,013 |
0.111 |
2,372 |
20,607 |
0.216 |
4,136 |
20,944 |
| GC-RMA MM threshold |
0.037 |
1,107 |
19,969 |
0.111 |
2,392 |
20,621 |
0.235 |
4,202 |
21,003 |
|
|
|||||||||
|
The number of false positives (FPs) and true positives (TPs) is for a given percentage of false positives out of the total number of positives (5%, 10% and 17% but percentages for each method are not exact due to P value granularity) and the corresponding P value cutoff for the MAS 5.0 present/absent calls. The calculations are based on a total of 60,624 true negatives (10,104 empty probesets in 6 replicates) and 23,436 true positives (3,906 probesets that can be aligned to transcripts in six replicates). The default P value cutoff is 0.060 for separating marginally present from absent transcripts. |
|||||||||
|
Schuster et al. Genome Biology 2007 8:R125 doi:10.1186/gb-2007-8-6-r125 |
|||||||||
