Table 2.
Comparison of computational methods
| Method | Validation set | PS ≥1a | PS ≥2b | Tsc |
|---|---|---|---|---|
| ESP Predictor | HeLa_2 (GeLC-MS) | 86% | 54% | 498d |
| STEPP13 | HeLa_2 (GeLC-MS) | 80% | 44% | 425d |
| Peptide sieve (PAGE-ESI)10 | HeLa_2 (GeLC-MS) | 77% | 43% | 413d |
| Peptide detectability12 | HeLa_2 (GeLC-MS) | 77% | 41% | 394d |
| ESP predictor | Plasma Hu14 SCX | 93% | 49% | 74d |
| Peptide sieve (MUDPIT-ESI) | Plasma Hu14 SCX | 82% | 46% | 65d |
| STEPP | Plasma Hu14 SCX | 69% | 36% | 51e |
| Peptide detectability | Plasma Hu14 SCX | 62% | 13% | 35f |
The ESP predictor demonstrates the best performance compared to existing computational methods. Refer to Table 1 for additional validation set information. STEPP, SVM technique for evaluating proteotypic peptides.
Protein sensitivity (PS): The percent of proteins with one or more peptides predicted by the ESP predictor to be among the five highest responding.
The percent of proteins with two or more peptides predicted by the ESP predictor to be among the five highest responding.
Test statistic (Ts). The sum of correct peptides among the five highest-responding peptides for all proteins in the validation set.
P < 0.0001
P = 0.0029
P = 0.6685 based on null distribution for the entire validation set, by permutation test.