News

This Masters’ degree project developed a predictive algorithm to categorize butter, butter spreads, and margarine/vegetable oil spreads according to their fatty acid profile, moisture, and total fat content based on the spectra collected by using handheld FT-NIR and portable FT-MIR devices. FT-NIR infrared and FT-MIR performances were similar, with a strong correlation (Rep >0.94) and low standard error of prediction for different analyzed parameters. SIMCA classification model based on FT-NIR and FT-MIR spectra effectively differentiated between butter, butter spread, and margarine/vegetable oil spreads.

The results were benchmarked against “classical” analysis.  Moisture and total fat content were determined using reference methods AOAC 920.116 and AOAC 938.06-1938, respectively. FA profile was determined using Gas Chromatography with flame ionization detector (GC-FID) (AOAC 996.06, 1996.). The FA profile showed that butter-containing products distinguished from margarine/vegetable oil spreads based on the presence of trans fats (TFA) (C18:1t) and butyric acid (C4:0).

The author concludes that portable FT-MIR and handheld FT-NIR technologies offer real-time and in situ analysis capabilities, enabling the dairy industry and regulatory agencies to make actionable decisions regarding FA, moisture, and total fat content and for nutrition, authentication, claims, and labeling purposes of these products.

The abstract and author contact details are available here.  The full text is being withheld until May 2026 at the author’s request. For an overview of FT-NIR see FAN's analytical techniques explainer for spectroscopy.

Photo by Sorin Gheorghita on Unsplash

In this study (purchase required) the researchers build a classification model for differentiate freshwater from seawater shrimp (prawns), Litopenaeus vannamei, based on fatty acid (FA) profiling in muscle and hepatopancreas.

They built an untargeted model, using k-nearest neighbor (KNN) and random forest (RF), to identify discriminatory variables.

They then identified, using orthogonal partial least squares-discriminant analysis (OPLS-DA) specific FAs to create their classification model: six (C22:6n3, C20:3n3, C17:0, C18:3n3, C20:5n3, and C20:2) from the muscle and seven (C22:6n3, C16:0, C18:3n3, C18:2n6, C20:2, C20:1, and C18:1n9) from the hepatopancreas.

They report that, using FA profiles from the two tissues, both KNN and RF had initial and cross-validated classification rates >93%, while the predictive classification rates of the models based on muscle FA profiles were higher than that of the models based on hepatopancreas FA profiles. They conclude, therefore, that FA profiles in muscle were more effective than hepatopancreas FAs for this promising classification method.

Photo by Dan Dennis on Unsplash