Table (V).
Per amino acid agreements for the three-state scenario
| TM | TR | SOL | avg | |||
|---|---|---|---|---|---|---|
| pred | UHS | |||||
| TM | 63 ± 11 | 29 ± 10 | 9 ± 6 | |||
| TR | 23 ± 7 | 44 ± 3 | 26 ± 4 | |||
| SOL | 13 ± 6 | 27 ± 9 | 64 ± 8 | |||
| 57 ± 3 | ||||||
| pred | WW | |||||
| TM | 36 | 14 | 2 | |||
| TR | 29 | 24 | 9 | |||
| SOL | 35 | 62 | 89 | |||
| 50 | ||||||
| pred | MHS | |||||
| TM | 71 ± 1 | 17 ± 4 | 5 ± 2 | |||
| TR | 19 ± 3 | 48 ± 1 | 30 ± 4 | |||
| SOL | 10 ± 2 | 35 ± 2 | 65 ± 2 | |||
| 61 ± 0 |
The table shows the percentage per amino acid agreements between the prediction and the PDB for the different regions for the UHS (with its standard deviation) in comparison to the Wimley & White scale. The performance of the MHS is also shown. The window size for averaging is 15 residues and (TM) represents the trans-membrane, (TR) the transition and (SOL) the soluble region. The percentages were calculated by dividing the correctly predicted number of amino acids by the total number of amino acids in that region. An average agreement (avg) was calculated by averaging the percentages of agreement for the diagonal elements of the matrix. While the average prediction agreement seems to be relatively low, note that there are three regions defined, so that the threshold between a good and a bad percentage of agreement would be 33% and not 50% as in the two-state system. For the Wimley & White scale both the octanol and the interface scale were used to establish a scale for three regions. The standard deviations for the UHS and MHS arise from cross-validation, whereas the scale of WW was tested on the whole dataset without cross-validation.