Table II. Decoy Quality Comparison Between the HMM Model17 and CRFSampler.
| Test proteins | HMMa | CRFSamplerb | CRFSamplerc | |||||
|---|---|---|---|---|---|---|---|---|
| Name, PDB code | L | α,β | Good (%) | Best (Å) | Good (%) | Best (Å) | Good (%) | Best (Å) |
| Protein A, 1FC2 | 43 | 2,0 | 9.59 | 2.7 | 20.9 | 2.08 | 24.8 | 2.09 |
| Homeodomain, 1ENH | 54 | 2,0 | 6.60 | 2.5 | 6.23 | 2.68 | 14.0 | 1.98 |
| Protein G, 2GB1 | 56 | 1,4 | 0.04 | 4.9 | 0.16 | 4.67 | 10.1 | 3.36 |
| Cro repressor, 2CRO | 65 | 5,0 | 0.46 | 3.9 | 1.94 | 4.05 | 13.3 | 2.37 |
| Protein L7/L12, 1CTF | 68 | 3,1 | 0.01 | 5.4 | 0.04 | 4.94 | 0.15 | 4.49 |
| Calbidin, 4ICB | 76 | 4,0 | 0.09 | 4.3 | 0.17 | 4.57 | 0.42 | 4.72 |
Only sequence information is used in both training and testing. In total, 100,000 decoys are generated by the HMM model.
Columns 1–3 list name and PDB code, length and number of α-helices and β-strands of the test proteins. Columns “Good” and “Best” list the percentage of good decoys (with RMSD ≤ 6 Å) and the RMSD of the best decoy, respectively.
a
Trained and tested using primary sequence and the results are taken from Ref. 17.
b
Trained and tested using primary sequence and 40,000 decoys are generated.
c
Trained and tested using PSI-BLAST sequence profile. Only 20,000 decoys are generated.