Rapid tannin profiling of tree fodders using untargeted mid-infrared spectroscopy and partial least squares regression

Table 2 Summary of calibration and prediction PLS-R parameters for the tannin profile

	LV	Regions (cm⁻¹)	R²C	RMSEC	RMSECV	R²P	RMSEP	RPD	RER	Cal. Bias
%CT (HBAI)	7	iPLS (1186–1454)	0.93	0.59	0.76	0.92	0.70	3.71	13.1	− 0.019
%CT (Thiolysis)	6	Main frequencies (551–1810/2700–3000)	0.87	0.89	1.08	0.88	0.93	2.80	11.5	− 0.215
mDP	6	Full range (550–4000)	0.75	0.64	0.84	0.66	0.77	1.86	7.16	− 0.015
PC	8	Fingerprint (700–1700)	0.97	4.96	6.82	0.96	5.60	5.26	14.1	0.208
cis	6	iPLS (9 regions^a)	0.96	6.74	8.39	0.95	7.07	4.24	13.3	− 0.157

LV latent variables, RMSEC/CV/P root-mean-square error or calibration/cross-validation/prediction, RPD ratio of prediction to deviation, RER range error ratio, %CT content of condensed tannins (%DM) measured by HCl-butanol-acetone-iron assay (HBAI) or thiolysis-HPLC (Thiolysis), mDP mean degree of polymerisation, PC % procyanidin units (% condensed tannins), cis % cis units (% condensed tannins)
^a646–673/790–846/877–904/1108–1135/1253–1280/1340–1367/1398–1425/1658–1685/1716–1743/1774–1810 cm⁻¹

ISSN: 1746-4811