Previously, we reported that near infrared reflectance spectroscopy (NIRS) can predict digestible amino acid (dAA) and nitrogen corrected true metabolizable energy (TMEN) content of ground solvent extracted (SE) and mechanically expelled (ME) soybean meal (SBM) samples with predictive accuracy comparable to the margin of error associated with rooster bioassays. These calibrations were developed on a Bruker Multipurpose Analyzer (MPA), a dual-channel FT-NIR instrument; however, in commercial settings such as feed mills, the single-channel TANGO-R is more commonly used, as it employs a more streamlined graphical user interface in comparison to the MPA. The current research assesses the practical implementation of our original calibrations by examining the influence of particle size and inter-instrument variation on predictive consistency. Spectra for 33 SE and 10 ME SBM samples were analyzed. Each sample was divided into two aliquots. One aliquot was ground using a 1093 Cyclotec Sample Mill (0.5-mm screen) and analyzed on a TANGO-R at a feed mill. The other aliquot was ground with a Retsch ZM200 centrifugal grinder (1.0-mm screen) and analyzed using our MPA, and the same TANGO-R was transported to our laboratory. Both instruments were equipped with calibration models established in our initial research. Statistical differences were evaluated using paired t tests. Significant differences (P < 0.05) were observed for TMEN and nearly all dAAs across grind size and between instruments, but mean differences were marginal (<5% of analyte average) and directionally consistent. Predictions from the 0.5-mm grind and TANGO-R were almost universally lower than their paired comparisons. These results indicate that grind size and inter-instrument variation exert minor, predictable effects on TMEN and dAA predictions. Since these differences follow a consistent direction rather than a random distribution, analyte-specific bias correction for instrument and/or particle size can be applied to the original calibrations. Consequently, NIRS could provide a reliable and transferable method for determining TMEN and dAA content of SE and ME SBM in poultry across Bruker FT-NIR platforms and sample preparation conditions.

Jones, T., C. Juzaitis-Boelter, C. Chen and A. Davis. 2026. Assessing calibration transferability and impact of particle size on near infrared reflectance spectroscopy predictions of nitrogen corrected true metabolizable energy and digestible amino acid content in solvent extracted and mechanically expelled soybean meal. 2026 International Poultry Scientific Forum, abstract M133.