Table 2.

Evaluation of secondary structure evaluation at t₀ and t₁₅ and for the global model on a 25-protein test set selected by Kennard-Stone algorithm

	α-helix			Ordered α-helix			Others			β-sheet			Anti//β-sheet
	RMSEKS	n. var	ζKS	RMSEKS	n. var	ζ KS	RMSEKS	n. var	ζ KS	RMSEKS	n. var	ζ KS	RMSEKS	n. var	ζ KS
PLS t0	7.13	5 LVs	2.96	7.62	5 LVs	2.73	12.22	3 LVs	1.24	5.36	4 LVs	2.82	6.12	7 LVS	2.48
PLS t15	6.63	6 LVs	3.18	7.44	6 LVs	2.8	9.3	5 LVs	1.63	7.07	3 LVs	2.14	6.52	8 LVs	2.33
PLS tGM	6.2	5 LVs	3.4	7.22	5 LVs	2.88	10.49	4 LVs	1.45	6.15	4 LVs	2.46	7.41	5 LVs	2.05
SVM t0	6.97		3.1	7.21		2.75	10.52		1.37	4.98		3.12	5.66		2.61
SVM t15	7.42		2.91	7.3		2.71	9.98		1.44	6.47		2.4	6.62		2.23
SVM tGM	7.72		2.8	7.47		2.65	12.42		1.16	6.63		2.35	6.94		2.13
ASLR t0	7.49	4 wnbs	2.82	7.9	4 wnbs	2.63	12.54	5 wnbs	1.21	5.25	4 wnbs	2.88	5.89	4 wnbs	2.58
ASLR t15	7.07	4 wnbs	2.98	8.07	4 wnbs	2.58	10.31	5 wnbs	1.47	5.66	4 wnbs	2.67	6.2	4 wnbs	2.45
ASLR tGM	6.83	4 wnbs	3.09	7.38	4 wnbs	2.82	11.12	5 wnbs	1.37	4.95	4 wnbs	3.06	6.1	4 wnbs	2.49
STDDEVREFKS	21.1			20.82			15.18			15.12			15.2

PLS, SVM and ASLR performances on a 25-protein test set selected by Kennard-Stone algorithm in the 1720–1480 cm⁻¹ spectral section at t₀, t₁₅ and for the global model (t_GM) including spectra recorded at t₀, t₁₅, t₁₀₅ and t_24h. RMSEKS, error in prediction (expressed in%), ζ^KS = STDDEV^REFKS/RMSEKS. STDDEV^REFKS is the standard deviation of the secondary structure content in test set selected by the Kennard-Stone algorithm. The optimal number of LVs (latent variables) for PLS and wvnbs (wavenumbers) for ASLR are reported