Time for this week’s QJART!
Linking Chemical Parameters to Sensory Panel Results through Neural Networks to Distinguish Olive Oil Quality
Cancilla JC, Wang SC, Diaz-Rodriguez P, Matute G, Cancilla JD, Flynn D, & Torrecilla JS
Journal of Agricultural and Food Chemistry 62 (2014) 10661-10665
Yum. And that’s all I have to say about that.
I don’t know many people who don’t love to dip a piece of fresh bread into some extra virgin olive oil. But would you be able to taste the difference between extra virgin and ordinary virgin olive oil? It’s okay, you don’t have to: Mathematics will do it for you!
In order to combat the growing trend of adulterated or falsely-labelled olive oils, researchers in Madrid, Spain have developed a method through which differences between extra virgin olive oil (EVOO), virgin olive oil (VOO), ordinary virgin olive oil (OVOO), and “lampante” olive oil (LOO, natural olive oil not fit for consumption). To do so, they used the combination of a sensory panel and the measurement of six chemical parameters of 220 olive oil samples, and then applied nonlinear mathematical modelling known as artificial neural networks (ANNs), which allow for the discovery of “nonlinear trends that exist between variables”.
First, the sensory panel were asked to evaluate the olive oils based on attributes considered desirable (green, ripe, and bitter) and related to defects (earthy, vinegar-like and muddy). They were also asked to grade the oils as EVOO or other.
Six chemical parameters were also measured in each oil; free fatty acid content (FFA, related to the acidity of the oils), peroxide value (PV, a measure of oxidation), two UV absorption parameters (K232 and K268), 1,2-diacylglycerol (DAG) content (a component found in a range of 1-3% in virgin olive oils), and pyropheophytin content (PPP, a degradation product of chlorophyll which is found in olive oils that have degraded through age or heat). The different graded olive oils provided different values for each of these tests.
Hmmm, they all look the same… but taste different and contain many of the same chemical components, but it different quantities!
ANNs were then used to link the results from both the chemical analyses and sensory evaluation, and through the identification of various relationships, were able to correctly classify (on average) 96% of olive oils. The researchers did note that while the ANNs used are have been successful in other food and chemistry-related scenarios, that the particular modelling used may not be as successful when looking at samples different to those used in this study.
Think about that the next time someone tells you that maths is useless after high school: it’s helping save you money at the supermarket every time you buy olive oil!