Combining sequence-based prediction methods and circular dichroism and infrared spectroscopic data to improve protein secondary structure determinations

BMC Bioinformatics

Table 2 RASP46 dataset cross-validation.

Method	α-helix (H)		β-Sheet (E)		Other (G, I, T, B, S, C)
	r	δ	r	δ	r	δ
CD	0.944	0.074	0.916	0.067	0.819	0.077
FTIR	0.940	0.076	0.900	0.076	0.740	0.070
CD_FTIR	0.955	0.066	0.938	0.060	0.858	0.068
SEQ	0.969	0.058	0.917	0.072	0.860	0.079
CD+SEQ	0.970	0.055	0.956	0.056	0.895	0.060
FTIR+SEQ	0.974	0.053	0.939	0.062	0.846	0.072
CD_FTIR+ SEQ	0.976	0.051	0.957	0.053	0.893	0.060

Cross-validation prediction accuracy of the secondary structure content prediction of proteins in the RASP46 dataset. The performance parameters from CD, FTIR, sequence prediction (SEQ), and consensus methods are shown. The best performance parameters for each secondary structural type are shown in bold.