Chemometric approach based on feed NIR spectra for rapid assessment of digestibility profiles in unconventional feedstuffs

Digestibility profiles serve as vital indicators of feedstuffs' quality, determining how effectively livestock utilise feed nutrients for growth and production. This study develops a rapid and non-destructive model for the digestibility quality of unconventional feedstuffs, which integrates feed NIR spectra with chemometric techniques. A total of 30 samples (citronella residues) obtained from the fermentation process were used, and their NIR spectra (1000~2500 nm) were acquired using a BUCHI NIRFlex N-500 spectrometer. Reference digestibility profiles, including dry matter (IVDMD), organic matter (OM), and pH of rumen fluid, were measured using the two-stage digestion method. Partial least squares regression (PLSR), ridge regression (Ridge), adaptive boosting (AdaBoost), and support vector machine regression (SVMR) models were then constructed to assess digestibility quality. Model performance was assessed using the coefficient of determination (R2), root-mean-square error (RMSE), and residual predictive deviation (RPD). The results show that the performance of the digestibility profile prediction model meets the accepted standards for NIRS calibration. Notably, the SVMR model exhibits exceptional stability and performance, achieving an R2 value of ≥ 0.97 and a low RMSE, outperforming the AdaBoost, Ridge, and PLSR models. These findings confirm that the calibration model has significant potential for further independent validation to serve as a rapid-quality assessment tool in analyzing the digestibility profiles of feedstuffs.