TABLE 3.
Single model standard errors of prediction for area-normalized spectra
RMS | Cross-valid lin | Direct valid | Frequencies in model | Model →% of the secondary structure | |
---|---|---|---|---|---|
α-Helix | 21.8 | 6.3 | 5.8 | – | −218.813 + 2.473 × A1545 + 0.565 × A1655 + 1.121 × A1611 |
β-Sheet | 17.6 | 7.6 | 7.0 | 1545–1655–1611 | 134.940 + 0.3861 × A1545 − 0.859 × A1655 − 0.741 × A1611 |
β-Turns | 4.2 | 3.6 | 3.2 | – | −14.043 + 0.475 × A1545 + 0.380 × A1655 − 0.650 × A1611 |
Random | 11.4 | 8.9 | 7.9 | – | 216.0146 − 1.285 × A1545 − 0.457 × (A1655) − 1.019 × A1611 |
For legend, see Table 1. A linear model was used for either direct or cross-validations. The standard error of prediction has been minimized for the four structures together, yielding the same three wavenumbers for all the models.