| Popis: |
238 samples, representing 19 vegetable feeds, were analysed for total and phytate phosphorus. Phytase activity was estimated in a limited number of feeds with an activity of >100 units/kg. Near infrared reflection spectra (NIRS) were taken between 1100 and 2500 nm in steps of 4 nm. By means of principal component analysis the feeds were grouped in starch- (n = 150) and protein-rich feeds (n = 88). NIRS-calibrations were developed using partial least square regression and tested by cross-validation. Total P varied from 0.22 to 1.25% and phytate P from 0.05 to 1.02% for high-starch feeds and, respectively, from 0.46 to 1.37% and from 0.22 to 0.56% for protein-rich feeds. Within these broad ranges, NIRS-values were highly correlated to measured total and phytate P with prediction errors of 0.08 and 0.08%, respectively, for high-starch feeds and 0.08 and 0.04%, for protein-rich feeds. However, considering the mostly small deviation of mean or calculated values for the individual feeds, NIRS is practical for predicting total P for wheat byproducts and phytate P for maize gluten feed. On the other hand, NIRS could be used for unknown or not tabulated products. Spectral signals from total and phytate P are partly based on direct responses from organic complexes but also on secondary relations through protein and fat. It was concluded that NIRS seemed not sensitive enough to detect phytase. |
