|
Number of periodic genes detected using different methods for simulated data |
|||
| P |
|||
|
|
|||
| Noise levels (σ) |
Missing ratio |
Lomb-Scargle method |
Our Algorithm |
|
|
|||
| 0.01 |
15% |
93 |
98 |
| 20% |
88 |
92 |
|
| 40% |
68 |
80 |
|
| 65% |
40 |
65 |
|
| 0.2 |
15% |
85 |
92 |
| 20% |
75 |
85 |
|
| 40% |
60 |
78 |
|
| 65% |
32 |
50 |
|
| 0.6 |
15% |
72 |
89 |
| 20% |
65 |
80 |
|
| 40% |
51 |
65 |
|
| 65% |
28 |
47 |
|
|
The simulated data consists of 1000 time series (genes). The total number of periodic genes is 100. P is the number of periodic genes that are statistically significant for a FDR level of q = 0.02. The number of false positives FP can be computed as FP = q*TP/(1-q), where TP is the number of true positives. Keeping q at a fixed level ensures that we do not sacrifice specificity for sensitivity. | |||
Liew et al. BMC Bioinformatics 2007 8:137 doi:10.1186/1471-2105-8-137 |
|||