Differentiation of organic and non-organic ewe's cheeses using main mineral composition or near infrared spectroscopy coupled to chemometric tools: A comparative study

• Published in: Talanta (Oxford) 2011-09, Vol.85 (4), p.1915-1919
• Main Authors: González-Martín, M. Inmaculada, Hernández-Hierro, José Miguel, Revilla, Isabel, Vivar-Quintana, Ana, González-Pérez, Claudio, García, Lorena Gómez, Riocerezo, Carlos Palacios, Ortega, Iris A. Lobos
• Format: Article
• Language: English
• Subjects: Analytical chemistry; Animals; Cheese - analysis; Chemistry; Discriminant Analysis; Exact sciences and technology; Female; Least-Squares Analysis; Main mineral composition; Minerals - chemistry; Near infrared spectroscopy; Organic Chemicals - analysis; Organic ewe cheese; Sheep; Spectrometric and optical methods; Spectrophotometry, Infrared - methods; Supervised pattern recognition methods
• ISSN: 0039-9140; 1873-3573
• DOI: 10.1016/j.talanta.2011.07.021

Two independent methodologies were investigated to achieve the differentiation of ewes’ cheeses from different systems of production (organic and non-organic). Eighty cheeses (40 organic and 40 non-organic) from two systems of production, two different breeds of ewe, different sizes, seasons (summer and winter) and ripening times up to 9 months were elaborated. Their mineral composition or the information provided by their spectra in the near infrared zone (NIR) coupled to chemometric tools were used in order to differentiate between organic and non-organic cheeses. Main mineral composition (Ca, K, Mg, Na and P) of cheeses and stepwise lineal discriminant analysis were used to develop a discriminant model. The results from canonical standardised coefficients indicated that the most important mineral was Mg (1.725) followed by P (0.764) and K (0.742). The percentage of correctly classified samples was 88% in internal validation and 90% in external validation, selecting Mg, K and P as variables.Spectral information in the NIR zone was used coupled to a discriminant analysis based on a regression by partial least squares in order to obtain a model which allowed a rate of samples correctly classified of 97% in internal validation and 85% in external validation.