Gabriella Andrássy-Baka

Identification of traditionally reared Mangalica pig's meat by near infrared spectroscopy using generalised partial least squares in open source R Project-a feasibility model study

The possibility for near infrared spectroscopy-based discrimination of meats originating from the extensively reared autochthonous breed of Mangalica and intensively reared commercial genotypes (Landrace, Large White, Landrace × Large White crossbreed) was investigated. Since there was a considerable difference between the intramuscular fat content of Mangalica and intensively-reared meats (average of 19.1 DM% vs 9.3 DM%, resp.), several sample selection options were applied to explore the impact of fat content on the results of NIR analysis. The system for discrimination was able to identify the different groups even when the discriminator equation was generated on very different samples and was tested on samples with overlapping fat content. The ratio of correctly classified samples was above 90% during cross-validation or for independent test samples of all comparisons, both in fresh or freeze-dried samples. Over 90% of independent fresh pork samples were correctly identified when the discriminator equation was generated with 70 randomly selected samples. This ratio increased up to over 95% when freeze-dried samples were applied. The generalised partial least squares package of open-source R Project seems to be a useful tool for qualitative analysis of NIR data recorded from meat samples.