Rapid prediction of chemical composition and degree of starch cook of multi-species aquafeeds by near infrared spectroscopy

Abstract

Ensuring aquafeeds meet the expected nutritional and physical specifications for a species is paramount in research and for the industry. This study aimed to examine the feasibility of predicting the proximate composition and starch gelatinisation (or cook) of aquaculture feeds (aquafeeds) regardless of their intended target species by near infrared (NIR) spectroscopy. Aquafeed samples used for nutrition experiments on various aquatic species with different nutritional requirements, as well as aquafeeds manufactured under varying extrusion conditions and steaming time to generate variable starch cook were used in this study. The various size pellets were ground before scanning by NIR spectroscopy, then models were developed to estimate dry matter, ash, total lipid, crude protein, and gross energy as well as starch cook. Proximate prediction models were successfully produced for diets with R2 values between 0.88 and 0.97 (standard error of cross-validation (SECV) 0.43 to 1.46, residual predictive deviation (RPD) 4.6 to 15.6), while starch cook models were produced with R2 values between 0.91 and 0.97 (SECV 3.60 to 5.76, RPD 1.2 to 1.9). The developed NIR models allow rapid monitoring of the nutritional composition, as well as starch cook, one of the major physical properties of aquafeeds. Models that provide rapid quality control assessment of diet characteristics is highly desirable in aquaculture research and the aquafeed industry.

Exploring the Analytical Complexities in Insect Powder Analysis Using Miniaturized NIR Spectroscopy

“…The use of RPD and RER values is not widely used, but some published papers for the prediction of macronutrients using benchtop NIR instruments report them. For instance, the prediction of micro- and macro-nutrients was carried out in soybean crops with RPD values ranging between 1.03 and 4.33 [ 38 ], or prediction of the chemical composition in aquafeed samples used for nutrition on various aquatic species reported RPD values mostly between 1.2 and 10.9, with only an excellent RPD value of 15.6 [ 39 ]. The reported RPD values in this manuscript (included in Table 2 ) are globally better (especially for the PLS models using SCiO) than the RPD values reported in the above-mentioned reviewed papers, especially considering that the models reported in this manuscript have been built using miniaturized NIR instruments.…”