2023 article
Data Analytics of a Global Nirs Database of Corn Nutrient Content
Oviedo-Rondon, E. O., Toscan, A., Fagundes, N. S., Vidal, J. K., Barbi, J., & Thiery, P. (2023, November 6). JOURNAL OF ANIMAL SCIENCE, Vol. 101, pp. 517–518.
Abstract Corn is the primary energy source for livestock and poultry. Corn nutrient composition and energy value vary due to genetics, country of origin, agronomic, harvest, and storage conditions. Analytical variability among and within-labs makes it more difficult to detect differences among sources and establish nutrient interrelationships to estimate amino acid (AA) content based on crude protein (CP), which is less variable. Near-infrared reflectance spectroscopy (NIRS) can standardize feedstuff evaluation and increase the sample size. The PNE system (Adisseo) has distinctive NIRS curves created using direct calibrations with in vivo data of standard chicken metabolizable energy (AME and AMEn) and AA digestibility studies. This study assessed the link between AA concentrations and CP and among proximate parameters (CP, crude fiber, ether extract, nitrogen-free extract, and ash), starch, AA concentrations and digestibilities with AME and AMEn. The PNE database used includes 84,331samples of maize scanned in diverse countries between 2018 and 2021 that came from Argentina (7,654), Brazil (49,672), Indonesia (18,072), Philippines (4,747), and the United States (4,186). Correlation and regression analyses between AA and CP were performed. The pairwise correlations (P < 0.001) observed between AA and CP ranged between 0.52 and 0.90. The positive linear associations with CP content can be used to predict the concentrations of all AAs in corn (P < 0.001). However, coefficients of determination (R2) ranged from 0.27 (Trp) to 0.80 (Val) or 0.81 (Phe). Prediction equations for Lys, sulfur AAs had lower R2 values (0.37 to 0.47). Although the country of origin and CP interacted (P < 0.001), R2 modestly increased when added to these models. The estimation of AME and AMEn based on proximate values, starch, amylose, amylose:amylopectin, AA concentrations, AA digestibility, and country of origin was done using multiple linear regression (MLR) and neural networks (NN) analysis. All factors were significant (P < 0.001), and MLR models including starch and amylopectin, had the best fit for AME (R2 = 0.87, AICc = 149,635) and AMEn (R2 = 0.87, AICc = 149,543). The NN models had a better fit for training and validation datasets for AME (R2 = 0.95, RASE = 15.18) and AMEn (R2 = 0.95, RASE = 15.23). The NN models helped to visualize the greater importance of starch, AA concentration (Trp, Ile, Thr), and Leu digestibility. In summary, some corn AA concentrations (Val, Phe, Ile) may be predicted from their CP content; other AA concentrations may be more difficult to predict. Corn AA content varies by country of origin and independent of CP content. Starch content, AA concentration, and AA digestibility are essential factors in determining energy value of corn. Relevant nutritional aspects were elucidated by data analytics of a NIRS database with direct calibration of in vivo parameters of energy utilization and AA digestibility.