Título:	Portable near-infrared spectroscopy and data fusion with in vitro digestibility improve prediction of indigestible neutral detergent fiber in tropical grasses.
Autoria:	OLIVEIRA, C. F. de ALVES, A. T. R. CARNEIRO, P. P. FURTADO, H. R. SILVA, J. dos S. CARNEIRO, S. C. MOURA, A. M. DETMANN, E. SILVA, T. E. da RUFINO, L. M. de A. RODRIGUES, J. P. P.
Afiliação:	CATARINA FERNANDES DE OLIVEIRA, UNIVERSIDADE FEDERAL RURAL DO RIO DE JANEIRO ARIEL THAIS RODRIGUES ALVES, UNIVERSIDADE FEDERAL RURAL DO RIO DE JANEIRO PRISCILA PEREIRA CARNEIRO, UNIVERSIDADE FEDERAL RURAL DO RIO DE JANEIRO HUGO REZENDE FURTADO, UNIVERSIDADE FEDERAL RURAL DO RIO DE JANEIRO JULIA DOS SANTOS SILVA, UNIVERSIDADE FEDERAL RURAL DO RIO DE JANEIRO STELLA CARDOSO CARNEIRO, UNIVERSIDADE FEDERAL RURAL DO RIO DE JANEIRO ANDRE MORAIS MOURA, UNIVERSIDADE FEDERAL RURAL DO RIO DE JANEIRO EDENIO DETMANN, UNIVERSIDADE FEDERAL DE VIÇOSA TADEU EDER DA SILVA, UNIVERSITY OF VERMONT LUANA MARTA DE ALMEIDA RUFINO, UNIVERSIDADE FEDERAL RURAL DO RIO DE JANEIRO JOAO PAULO PACHECO RODRIGUES, CNPGL.
Ano de publicação:	2026
Referência:	Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, v. 363, pt. 2, 128408, 2026.
Conteúdo:	Indigestible neutral detergent fiber (iNDF) is a relevant nutritional fraction related to fiber digestion in ruminants and is widely used in intake modeling. However, its estimation still depends on long time in situ incubations, which involve labor, high costs, and ethical constraints. We evaluated the predictive performance of iNDF in tropical grasses using portable near-infrared (NIR) spectroscopy alone or combined with laboratory composition and in vitro NDF digestibility (IVNDFD). A total of 275 samples of Urochloa and Megathyrsus spp. were collected from three regions and analyzed for dry matter (DM), organic matter (OM), nitrogen (N), neutral detergent fiber (NDF), IVNDFD, and iNDF. Spectra were acquired from dried and ground samples using a portable reflectance NIR spectrometer operating from 900 to 1700 nm. Five predictor sets were evaluated: spectra only, spectra plus laboratory composition (DM, OM, N, and NDF), spectra plus IVNDFD, spectra plus laboratory composition and IVNDFD, and a non-spectral model with IVNDFD and laboratory analyses (DL). The dataset was divided into train (80%) and test (20%) subsets using the Kennard-Stone algorithm, stratified by genus. Forty-nine spectral preprocessing combinations were tested, and partial least squares regression models were fitted in the train set. In external validation, spectra plus IVNDFD, and spectra plus laboratory composition and IVNDFD showed the best performance, with RMSEv values of 26.8 and 27.1 g kg− 1 DM, R2 v values of 0.888 and 0.889, and RPDv values of 2.96 and 2.97, respectively, whereas spectra only and spectra plus laboratory composition remained below an RPDv of 2.0. Bootstrap comparisons supported the superior and more stable performance of spectra plus IVNDFD, and spectra plus laboratory composition and IVNDFD. Integrating portable NIR spectra with IVNDFD improved iNDF prediction and offers a promising alternative to routine reliance on long in situ procedures.
Thesagro:	Forragem Planta Forrageira Análise de Laboratório Digestibilidade In Vitro
NAL Thesaurus:	Chemometrics
Palavras-chave:	Quimiometria Espectroscopia
Digital Object Identifier:	https://doi.org/10.1016/j.saa.2026.128408
Tipo do material:	Artigo de periódico
Acesso:	openAccess
Aparece nas coleções:	Artigo em periódico indexado (CNPGL)