News

Authentication of Extra Virgin Olive Oil (EVOO) sometimes requires a panel of different tests and – with more sophisticated adulteration – a weight of evidence interpretation.  For more crude adulterations a single test is often enough.

One of the available tests is for fatty acids ethyl esters (FAEE).  These are more concentrated in lower quality oils (e.g.improperly stored or overripe), formed from ethanol which is a result of fermentation. EU legislation specifies a maximum 35 mg per kg FAEE concentration in EVOO.

FAEE concentration is officially measured using gas chromatography (GC) after recovery by silica gel column chromatography. While highly accurate, this method is complex, time-consuming, and relatively expensive.

This paper (purchase required) reports an alternative approach to FAEE measurement by using infra-red spectroscopy (FT-IR) with machine learning. A dataset of 170 olive oil samples with FAEE concentrations ranging from 1.81 mg/kg to 109.00 mg/kg were analysed using FTIR. Spectral data were preprocessed and used to train various regression models.

The authors report that the best performance was obtained with an XGBoost model. Explainable AI techniques (SHAP) enabled interpretation of the model and identification of spectral regions mostly associated with FAEE content.

They conclude that combining FT-IR spectroscopy with advanced ML models—particularly XGBoost—can effectively predict the concentration of FAEE.

Photo by Massimo Adami on Unsplash

This paper (open access) reports the development of a classification model for the geographic origin on black tea based upon measuring a panel of 15 trace elements by X-ray fluorescence (XRF).  XRF is a non-destructive technique.  The only sample preparation required is grinding the tea leaves into a fine powder.

The model could discriminate between 10 major tea-producing regions.  It was built using reference samples obtained, via tea industry contacts, directly from plantations or primary processing facilities.   791 black tea samples were collected in total: Assam (272 samples), Burundi (40 samples), Darjeeling (145 samples), Ethiopia (40 samples), Keemun (115 samples), Kenya region 1 (41 samples), Kenya region 2 (40 samples), Malawi (40 samples), Rwanda (10 samples), and Sri Lanka (48 samples).

Two unsupervised analysis techniques were used to visualize high-dimensional data, and six supervised models were employed to discriminate the ten GI regions.

The authors report that machine learning models, including random forest, support vector machine, k-nearest neighbours, linear discriminate analysis, and the deep learning multilayer perceptron (MLP) model, demonstrated superior predictive capabilities compared to the traditional partial least squares discriminant analysis model. The MLP model achieved the highest performance, with a 97.7 % overall F1 score in predicting the geographical origins of 532 authentic samples across ten GI regions.

The authors also Identified Rb, Sr, Mn, Si, and Cl as geographical markers for African region discrimination.

The conclude that their work could form the basis and foundation for an international database of tea Geographic Origin, enabling cheap and quick authenticity verification testing.

Photo by Oleg Guijinsky on Unsplash

Most test methods and research into the authenticity of edible oils are focussed on differentiating different plant species or on different grades of olive oil.  There has been relatively little focus on different grades of sunflower oil.  Commercial sunflower oil is sold as three different grades with increasing price premium; standard Sunflower Oil (SFO), Medium Oleic Acid (MOSFO) and High Oleic Acid (HOFSO).  HOFSO is more stable to repeated heating/cooling cycles and so is the grade typically required for fast food restaurants.  It is also available as a premium product sold direct to consumers.

In this paper (open access) the researchers used Spatially Offset Raman Spectrocopy (SORS, a portable non-invasive sensor) to build statistical models that could differentiate HOFSO from those that were not HOFSO (i.e. either MOSFO or SFO).  Although the reference samples used to build the model were purchased from commercial outlets rather than being of verified authenticity, the fact that two different unsupervised mathematical plus a number of supervised approaches all led to similar classification models, and that the models were validated with samples independent of the training sets, gave increased confidence in the model.

The authors conclude that the use of  SORS in combination with the developed chemometric models is an effective tool for the HOSFO authentication. The approach is simple and rapid, with instrumental fingerprints from portable analyser in less than 2 min and without requiring sample preparation.  This approach would class as Green Analytical Chemistry.

Image from the paper

In this study (open access) a non-targeted method of headspace-solid phase microextraction with gas chromatography coupled to mass spectrometry (HS-SPME-GC–MS) was developed to achieve the characterization, classification, and authentication of different coffee samples according to geographical production region, and variety (arabica/robusta). Moreover, decaffeinated and non-decaffeinated instant coffee samples were analyzed. Some samples of chicory, a potential coffee adulterant, were also been included. The GC–MS fingerprints were used to classify and characterise the analyzed coffees using principal component analysis (PCA), partial least squares-discriminant analysis (PLS-DA) and partial least squares (PLS) regression.

185 reference samples were used.  42 were chicory (a typical coffee substitute or adulterant), 96 were coffee from three geographical production regions (Vietnam, Cambodia, and Costa Rica) and species (Arabica, Robusta, and Arabica-Robusta mixture), and 47 samples were soluble coffee (decaffeinated and non-decaffeinated). The chicory samples were purchased from Barcelona supermarkets and the coffee samples were from Vietnam, Cambodia and Costa Rica local supermarkets. Each paired PLS-DA model was built using 70 % of samples randomly selected from each group as the calibration set while the remaining 30 % of the samples were employed as the prediction set.  The authors compared models generated using two different GC columns and operating conditions.

The authors tested their model on mixtures prepared in-house from the same reference set:: Vietnamese Arabica Coffee adulterated with Vietnamese Robusta Coffee and Vietnamese Robusta Coffee adulterated with chicory.  They reported that the model could classify adulteration levels down to 15%.

(image from the paper)

This paper (purchase required) reports the use of a portable optical sensor (Multi-Spectral Imaging) to build a classification model for detecting milk adulteration. This encompassed mixtures of milk from different species (cow, goat, and sheep), as well as dilution of cow’s milk with water. The study's scope also included milk with diverse heat treatments, fat content, and commercial brands.

The authors report that discriminant analyses provided reliable predictive models, with Accuracy and Cohen's Kappa values ranging between 0.80 and 1. In quantitative studies, the quantification of milk mixtures at a minimum percentage interval of 10% was detected with Mean Absolute Error (MAE) values between 0.14 and 0.05, and 0.03 for cow's milk adulterated with water at adulteration levels of 5%.

The authors conclude that the portability of these instruments adds a significant advantage by enabling on-site and real-time determination and quantification of milk adulteration.

This study (open-access author’s link available until February 14, with thanks to Michele Suman for sharing) reports the development and validation of a non-targeted classification method for authenticity of dried oregano leaves by atmospheric pressure matrix-assisted laser desorption ionization mass spectrometry (AP-MALDI-MS).

The model was trained on 23 authentic oregano samples (sourced from a reputable company with full supply chain traceability - originated from Italy, France, Turkey, or Albania, harvested between 2019 and 2022) along with five pure adulterants (dried leaves of savory (Satureja montana), myrtle (Lagerstroemia indica), sumac leaves (Rhus coriaria), strawberry tree (Arbutus unedo), and olive tree (Olea europaea)), plus sixteen adulterated oregano samples, intentionally mixed with the above mentioned adulterants at ranges between 5 % and 60 %.

The most abundant signals were characterized by collision induced dissociation and library search, the spectral data were submitted to statistical analysis. A basal inquiry of the data by partial least squared discriminant analysis (PLS-DA) was carried out for the simple assessment of the discrimination capabilities of the ± AP-MALDI-MS signatures. The researchers then constructed two distinct random forest (RF) classifiers using the positive and negative most informative ions teased out by recursive feature elimination from the training sets. The aforementioned most significant variables (m/z values) were also merged by mid-level data fusion and used to build a third RF classifier.

They report that the cross-validations of the three RF classifiers achieved good outcomes as demonstrated by the satisfactory values of overall accuracy (84.9 %, 92.1 %, and 92.8 %, respectively). The three RF classifiers were tested on the hold-out data, which revealed reliable classifier performances (accuracy 80.1 %, 87.0 %, and 85.4 %).

Photo by 360floralflaves on Unsplash

This FSA project develops a conceptual framework for modelling and capturing the full range of costs attributed to food crime on UK society, along with an assessment of the availability of data that would be necessary to produce this model.

An economic framework was developed for estimating the economic cost of food crime which uses:

1. Victim costs: Direct economic losses suffered by crime victims, including medical care costs and lost earnings.
2. Criminal justice system costs: Costs of anti-food crime activities, legal and adjudication services, and corrections programs including incarceration.
3. Crime career costs: Opportunity costs associated with the criminal’s choice to engage in illegal rather than legal and productive activities.
4. Intangible costs: Indirect losses suffered by crime victims, including pain and suffering, decreased quality of life, and psychological distress.
5. Market costs: Loss of profits for genuine firms.

Analysis was also conducted to assess how these costs can be calculated given available data sources. Finally, an assessment of the possibility of applying machine learning or other tools to build algorithms to calculate the costs was carried out.

The findings of this project will eventually be used in a phase 2 of the work which will look to build a model to provide preliminary estimates of the cost of food crime to UK society. 

Read full report.