ARS | Publication request: Determination of lipid content in poultry feeds by partial least squares models with discrete cosine transform

Submitted to: Meeting Abstract
Publication Type: Abstract Only
Publication Acceptance Date: July 9, 2010
Publication Date: October 18, 2010
Citation: Holser, R.A. 2010. Determination of lipid content in poultry feeds by partial least squares models with discrete cosine transform; Abstract #237758. Twelfth International Conference on Chemometrics in Analytical Chemistry, 18-21 October, 2010, Antwerp, Belgium. Meeting Abstract.

Interpretive Summary: The predictive capability of a partial least squares model to estimate the amount and type of lipid compounds in poultry feeds was tested using commercial feeds supplemented with docosahexaenate, the ester of DHA. Poultry feeds can be formulated with different amounts of saturated or unsaturated fatty compounds which influences the nutritional quality of the product. DHA is a highly unsaturated fatty acid that has shown numerous health benefits in humans and when fed to poultry it is stored in the poultry meat and eggs. Near infrared spectra were collected over the range 10 000 – 4 000 cm-1from samples of commercial starter, grower, and finisher feeds treated with five concentration levels of DHA in a factorial experimental design. These diffuse reflectance spectra were processed by multiplicative scatter correction (MSC), first derivative, second derivative, or discrete cosine transform (DCT). DCT was used as a variable selection technique. Partial least squares models developed from these processed spectra were compared to a model developed with unprocessed spectra. The results demonstrated a reduction in RMSEP values for all models developed using processed spectra with the DCT transformed spectra providing the lowest values compared to unprocessed spectra.

Technical Abstract: N/A