Determination of carbohydrate and lignin content in feedstuffs for monogastric animals using near-infrared spectroscopy

Abstract

Carbohydrates (CHO) are the principal constituents of the diets given to monogastric animals around the world. They supply the largest part of the energy to the animal, despite the fact only monosaccharides, disaccharides and starch can be enzymatically broken down in the small intestine into monosaccharides that are readily absorbed. In contrast, oligosaccharides (OS), resistant starch, non-starch polysaccharides (NSP) and lignin resist enzymatic digestion and thus cannot be absorbed directly. They compose the so-called dietary fiber. The soluble fraction of fiber is almost completely fermentable and causes benefits to the gastrointestinal tract while the insoluble fibre is less degraded and therefore lead to a decrease in nutrient and energy digestibility. The CHO fraction and its composition is highly variable among feed ingredients but also amongst the different varieties of the same ingredient. To improve the nutrient utilization and to formulate animal diets that are more efficient it is necessary to characterize the CHO fraction of the ingredients. In addition, with the growing use of supplementary enzymes, such as xylanase, β-glucanase or mannanase to mitigate the anti-nutritive effect of fiber the determination of the CHO components is essential. To make a full account of the CHO and lignin constituents, several wet chemical enzymatic, chromatographic and gravimetric methods need to be employed, which are expensive and time-consuming. Near-infrared spectroscopy (NIRS) overcomes these limitations because it is a rapid, nondestructive, multi-parametric technique and requires minimal sample preparation. In this work, around 300 samples of animal feedstuffs, consisting of cereals, cereal co-products, protein-rich and fiber-rich feedstuffs were used to develop and validate near-infrared (NIR) partial least square regression (PLSR) models. Spectral pretreatment optimization, sample selection and outlier detection were considered for each model. Robust predictions (R2 P>0.90) were obtained for digestible carbohydrates (sugars and starch) and for most of the total or insoluble components of dietary fiber (lignin, NSP, cellulose, β-glucan, non-cellulosic polysaccharides (NCP), arabinose, xylose, galactose, rhamnose and uronic acids). The OS, fructans were well predicted in cereals (R2 P=0.94) and linearity (R2 CV>0.70) was also found for α-galactosides (raffinose, stachyose and verbascose) in protein-rich feedstuffs. The obtained results were in general very satisfactory and demonstrate that NIRS is a reliable tool to characterize the carbohydrate fraction of a variety of feedstuffs.