Technical note: comparison of Raman, mid, and near infrared spectroscopy for predicting the amino acid content in animal meals.

The objective of this study was to compare three infrared spectroscopy techniques for routine evaluation of AA in animal meals. Animal meals (n = 54) with known AA contents were scanned with a near (NIRS), mid (FTIR), and Raman infrared spectrometer. For NIRS and Raman, samples were scanned "as is", whereas for FTIR, samples had to be finely ground before scanning to obtain reasonable spectra. Both FTIR and Raman data suffered from noise; for Raman, this prevented the development of calibrations. Using derivatized spectral data and a standardized outlier removal procedure, calibrations for nutritionally relevant AA could be developed that were equivalent for both NIRS and FTIR. The variation across AA tested explained (r2) by these calibrations was 70% for NIRS and 68 + 3% for FTIR. Removing spectral data between 4,000 and 2,000 cm(-1) from the FTIR data improved calibrations (P = 0.09) and explained an average of 77% of the variation with prediction errors lower than obtained with NIRS (P < 0.01). However, FTIR calibrations based on the entire or the shortened spectrum contained fewer samples than did NIRS calibrations (41 and 39 vs. 48, respectively; P < 0.01) because more samples were removed as outliers. In conclusion, Raman did not yield acceptable spectra for animal meals. For FTIR, sample preparation was more time-consuming because the samples required grinding before analysis. Using the entire mid-infrared range, FTIR calibrations were comparable to NIRS calibrations. Calibrations for FTIR were improved by eliminating wave numbers that exhibited more noise, resulting in prediction errors better than those for NIRS. Thus, FTIR has the potential to yield better calibrations for AA in animal meals than NIRS, but it requires greater care in sample preparation and scanning.