The main risk factors associated with an inadequate nutritional balance are dyslipidemia, high blood pressure, diabetes, hypercholesterolemia, obesity, sedentary lifestyle, etc. The prevalence of hypertension and dyslipidemia is the highest in Argentina compared to the rest of the Latin American countries [1]. Both health conditions mostly contribute to cardiovascular morbidity and mortality [2]. Nowadays, Argentine ranks fourth in cardiovascular mortality in Latin America, which began in developed countries, but is currently spreading to the world population [3,4].

The “westernization” of daily diet has led to increase in the consumption of industrialized foods, which are increasingly more appetizing, practical, and easy to prepare, although the main drawback of these foods is not providing complete, harmonious and healthy nutrition. In order to overcome this problem, it is necessary to consider different conditions on the fatty acids (FAs) content of vegetable oils (VOs): i. the presence of saturated FAs, which is the main dietary determinant of blood cholesterol level; ii. the content of trans-FAs, since their consumption would cause an increase in plasma lipids; iii. the content of ω3 and ω6 polyunsaturated FAs, where fats must provide 25 %–35 % of the total caloric value, and must also be distributed in 10 % or less as saturated fatty acids, and 20 or 25 % as unsaturated fats (ω9: 20 %; ω6: 5 %; ω3: 1 %). The ideal ω6:ω3 ratio should be 4:1 to 10:1 [[5], [6], [7]].

Undoubtedly, dietary unsaturated FAs and their metabolites not only provide health benefits, especially for cardiovascular diseases [8], but also in psychiatric disorders [[9], [10], [11]], depressive symptoms [[12], [13], [14], [15], [16], [17]], neurodegenerative diseases [[18], [19], [20], [21]], and in central nervous system [22,23], cognitive function and mood also in healthy adults [24], aging [25,26], muscle strength and performance [27], human intestinal microbiota recomposition [28], inflammatory processes in general [[29], [30], [31]], regulation of cell structure and neuronal or synaptic function contributing to ingestive behavioral outcomes [32], and blood-brain barrier integrity and glymphatic function [33], as demonstrated in animal tests and clinical trials.

The analysis of some of the physical and chemical characteristics of fats and oils is mandatory since their properties derive from them. In simple products, this characterization allows establishing counterfeits and identifying new products. In routine analysis, determinations of iodine and saponification values, together with other tests and qualitative tests for adulteration, are enough to confirm the identity and safety of most fats and oils.

The iodine value (IV) is a measure of the total number of double bonds in fat and oil compounds. It is expressed as the number of iodine grams that react with double bonds in 100 g of fat or oil. The determination is carried out by dissolving a weighed sample in a non-polar solvent such as cyclohexane. The double bonds react with an excess of a solution of iodine monochloride in glacial acetic acid (Wij's reagent). After the reaction (speeded up with mercury ions), excess iodine monochloride is decomposed into iodine by adding an aqueous solution of potassium iodide which is then titrated with a standard solution of sodium thiosulfate.

The saponification index (SI) measures the chain length or average molecular weight (AMW) of all FAs present in the sample as triglycerides. The higher the saponification index, the shorter the average length of the fatty acids and the AMW of the triglycerides that make up the fat, and vice versa. The SI is expressed as the number of milligrams of KOH required to saponify the FAs present in 1 g of fat. Taking into account that in the saponification reaction 1 mol of fat or oil reacts with 3 mol of KOH, each mole of fat consumes 168,000 mg of KOH, therefore, SI = 168,000/AMW. That is, the saponification index of a fat is inversely proportional to its average molecular weight.

It is important to highlight that these methods for determining the iodine value and the saponification index are considered the official ones, although common electrochemical methods such as Gas Chromatography coupled with a mass spectrometer as detector is able to replace the previous stoichiometric ones, due to speed and precision in the delivery of results. Nowadays, industry and laboratories rely heavily on fatty acid profiles as a tool for quantification, development and counterfeit detection.

In the context of contemporary data prediction, Machine Learning (ML) is a field of artificial intelligence (AI) that enables computer systems to learn and improve from data without being explicitly programmed. It involves a set of algorithms that can identify patterns, make predictions, and make decisions autonomously, as these are feed with large amounts of data. ML is currently being used in the vegetable oil industry for various applications, including identifying oil types, detecting adulteration, and analyzing oil quality. ML models can achieve high accuracy in classifying oils and predicting adulteration levels, even surpassing traditional chemical methods. Furthermore, ML can be used to analyze spectroscopic data like Raman and hyperspectral images to assess oil quality and identify different oils [34,35].

Nevertheless, ML algorithms has seen limited applications in solving food science related problems, coupled with advanced food analysis equipment. To the best of our knowledge, the rapid validation of a variety of edible oil quality by coupling Raman spectroscopy with machine learning has not been previously demonstrated.

Among the wide range of possible applications of ML, it also involves attempts for improving results found within the Quantitative Structure-Property Relationships (QSPR) Theory [[36], [37], [38]], which predicts the physicochemical properties of chemical compounds based on their molecular structures. The QSPR semiempirical parallelisms are particularly useful when it is necessary to predict unknown experimental properties. QSPR studies applied to the prediction of properties of complex chemical mixtures, such as natural oils, are a relatively new and unexplored field [[39], [40], [41], [42]].

QSPR analyzes are based on molecular descriptors [[43], [44], [45]], that is, numerical variables containing structural information about the constitutional, topological, geometric, hydrophobic and/or electronic characteristics of the studied molecules. Molecular descriptors are the crucial variables that determine the success of any QSPR study. In multicomponent systems, mixture descriptors [39,41] can be defined based on the descriptors of the FA components and their weight percentage compositions in the mixture.

It is known that the qualitative and quantitative compositions are perfectly defined in synthetic mixtures of few components, and so the mixture descriptors are straightforward calculated. This is not the case for natural oils, where the composition determination is not trivial. The experimental analysis of natural products allows the identification and quantification of most mixture components, although not all of them.

Due to the importance of vegetable oils in the organoleptic and nutritional properties conferred on foods, the present study formulates two variable-QSPR models for the first time in order to typify and characterize different plant-derived VOs. Therefore, the ratio between the saponification and iodine variables, p=SI/IV, is used for prediction. This proposed index combines both the average length and the AMW of the VOs with the total number of double bonds. High p-values suggest high SI and low IV indices, and vice versa. Non-conformational molecular descriptors are employed, in other words, descriptors that are completely based on the constitutional and topological characteristics of VOs.

The set of 144 VOs considered here contains up to 8 FA components in different compositions, such as lauric acid (12:0), myristic acid (14:0), palmitic acid (16:0), stearic acid (18:0), arachidic acid (20:0), oleic acid (18:1, ω9), linoleic acid (18:2, ω6), and α-linolenic acid (18:3, ω3). The molecular structures of these FAs are shown in Fig. 1 and, as can be appreciated, these involve carboxylic acids of different chain lenght.