Novel methods for secondary structure determination using low wavelength (VUV) circular dichroism spectroscopic data

Table 1 The cross-validation performance of various algorithms using the SP175 reference dataset [9] with the standard [1] secondary structure assignment scheme.

Dataset	Structure	SELMAT3		SELMAT1_norm		PLS		PLS-opt
		δ	r	δ	r	δ	r	δ	r
SP175	α _R	0.048	0.956	0.046	0.960	0.040	0.971	0.041	0.970
	α _D	0.035	0.809	0.035	0.811	0.036	0.791	0.037	0.779
	β _R	0.073	0.792	0.064	0.849	0.063	0.853	0.059	0.870
	β _D	0.020	0.913	0.019	0.921	0.023	0.889	0.025	0.867
	turn	0.052	0.325	0.053	0.297	0.052	0.332	0.051	0.319
	other	0.050	0.717	0.046	0.770	0.050	0.720	0.045	0.771
SP175 (nr)	α _R	0.049	0.954	0.048	0.956	0.041	0.970	0.042	0.969
	α _D	0.037	0.776	0.036	0.790	0.037	0.778	0.038	0.764
	β _R	0.083	0.725	0.067	0.832	0.065	0.841	0.061	0.862
	β _D	0.023	0.891	0.021	0.902	0.024	0.880	0.026	0.857
	turn	0.055	0.261	0.054	0.277	0.053	0.302	0.052	0.295
	other	0.055	0.671	0.047	0.754	0.054	0.683	0.046	0.764

The (nr) tag indicates that the cross-validation was carried out under more stringent (non-redundant) conditions such that no proteins in the training set with the same CATH homologous superfamily as that of the test protein were included. The best results (lowest δ or highest r) for each secondary structure type with the SP175 and SP175(nr) datasets are shown in bold.

ISSN: 1471-2105