Authentication of the geographical origin and the botanical variety of avocados using liquid chromatography fingerprinting and deep learning methods

https://doi.org/10.1016/j.chemolab.2020.103960Get rights and content

Highlights

  • Obtaining of the normal-phase lipid chromatographic fingerprint of avocado fruits.

  • Development and validation of multivariate classification methods.

  • Avocado fruits are differentiated regarding both geographical origin and cultivar.

  • Specific performance features of classification methods has been applied.

Abstract

The lipid chromatographic fingerprint of different avocado fruits have been acquired and two classification multivariate methods, partial least squares-discriminant analysis (PLS-DA) and support vector machine (SVM), have been successfully tested in order to discriminate and classify a higher variability of avocado samples. Two authentication goals have been achieved attending to: (i) the geographical origin, and (ii) the botanical variety or cultivar. However, to our knowledge, there are no antecedents aimed at comparing and classifying avocado fruits. The pulp oil fraction of the avocado fruit was first extracted using pressurised liquid extraction from the previously lyophilised pulp. Then the 190–400 ​nm UV-absorption fingerprints were obtained from the avocado oils using normal phase high performance liquid chromatography coupled to an absorption diode-array detector ((NP)HPLC-DAD) and the 220 ​nm spectra were then selected for classification model building. Several input-class classification strategies were applied and the classification models were externally validated from the specific success/error contingencies. In addition, some quality metrics, i.e. sensitivity (or recall), specificity, precision, negative predictive values, efficiency (or accuracy), AUC (area under the receiver operating curve), Mathews correlation coefficient and Kappa coefficient, were determined to evaluate the performance of each classification model (PLS-DA and SVM) and the results clearly show that SVM method is the most proficient.

Introduction

Avocado (Persea americana Mill., Lauraceae) is an ever-green tree originating from Mesoamerica and introduced in southern Spain during the sixteenth century by the Spaniards. It is botanically classified into three groups, which have been termed the Mexican (Persea americana var. drymifolia), Guatemalan (Persea nubigena var. guatemalensis) and West Indian (Persea americana var. americana) races. The differences between them are basically of morphological, physiological and horticultural traits. Each race has unique ecological adaptations and identifiable characteristics. Most commercial avocado cultivars are interracial hybrids, developed from chance seedlings, with different degrees of hybridization. The most well-known and marketed varieties are the Hass and Fuerte varieties [1].

The approximate composition of avocado pulp has been reported by the USDA (United States Department of Agriculture) National Nutrient Database for Standard Reference as follows: water, 73.23%; total lipids, 14.66%; proteins, 2%; dietary fibre, 6.70%; sugars, 0.66%; and carbohydrates (by difference) 8.53% [2].

Despite the origin of this crop is Central America, avocado has a high nutritional content that has recently aroused increasing global interest and it is now dispersed worldwide in tropical and subtropical regions. In 2017, the world avocado production was almost 6 million of tons. The leading countries in avocado production were Mexico (more than 2 million tons), Dominican Republic, Kenya or Chile among others. American continent conquers the avocado production (74%) and also have the highest number of harvested hectares (406,464 ​ha), followed by Africa (11.7%), Asia (11.2%) and Europe (1.6%) [3]. Spain is a special case in avocado cultivation since it is the only European country with a significant commercial production (about 90,000 tons in 2017). As more countries produce avocado for export or become familiar with the fruit, the increase in production area, production and consumption is likely to continue. Avocado marketing threshold value depends on the cultivar [4]. Moreover, main countries producers of avocado as Mexico or Peru with cheap labour, looser environmental regulations, and large acreage planted in avocados, seemingly are able to undercut the domestic fruit prices, providing consumers with cheaper priced avocados. This fact evidences the importance of knowing the traceability of avocado fruits.

Each vegetable species has a characteristic lipid profile. The fatty acid profile of avocado oil has been relatively consistent among studies, but the relative concentration of each fatty acid component was found to vary considerably. However, the oleic (42–51%) and palmitic (20–25%) fatty acids are always present in a larger proportion in avocado fruits since the predominant triacylglycerols are OOO (21–34%) and OOP (19–24%), where O and P denote oleic and palmitic acids, respectively [5]. There are many reports on the lipids contained in avocado fruit [6]. Moreover, some of them have proven that the oil content and composition vary according to the location of the orchard, the variety, the number of days between flowering and harvest, between others [7]. Several works have studied and quantified the fatty acid profile of different avocado varieties collected from different geographical regions. Nevertheless, all these studies are based on the quantification of specific compounds, such as oleic acid or tri-unsaturated fatty acids [8,9].

Fingerprinting is described as a battery of analytical techniques or methods based on treating the entire or a part of the chromatogram as a whole, without identifying or quantifying each compound [10]. The fingerprinting methodology uses nonspecific signals, where all the information implicitly found in the samples is used in a non-selective way with the main aim of characterizing or authenticating the food. Fingerprinting is based on obtaining as much useful information of the sample as possible without necessarily identifying or quantifying the compounds present.

The relevance of fingerprinting approach is undeniable too; they offer attractive advantages over other strategies, regarding analysis simplicity, accuracy and rapidity. The most widely used analytical techniques in fingerprinting analysis for the authentication and detection of adulteration of vegetable oils are infrared (IR) [11] and nuclear magnetic resonance (NMR) [12] spectrometries and both gas and liquid chromatographies coupled to numerous measuring devices [10]. Regardless of the approach selected, the obtained data require to be treated using appropriate chemometric tools to extract the unspecific and non-evident information of interest that is implicitly contained on the data. Focusing on chromatographic fingerprint, signal coming directly from the chromatographic instrument is processed and treated as a whole. Some examples of pattern recognition learning algorithms are unsupervised methods, and supervised ones [13,14].

In a previous paper it was displayed the discrimination of three avocado varieties using the entire lipid chromatographic profile (mainly related to the triacylglycerol profile) of the fruit applying the fingerprinting methodology [15]. Both reverse and normal phase liquid chromatographic coupled to diode array detector fingerprints were recorded and different multivariate classification strategies were tested. Normal phase conditions were proved to be better than reverse phase conditions for this purpose, showing shorter analysis time and better performance of the classification models. The PLS-DA method showed the best classification results.

To our knowledge, there are no other antecedents describing the identification, comparison and/or classification of avocado fruits according to both their botanical and geographical origin using the chromatographic fingerprinting from the fat fraction merging with chemometric tools.

In this paper, experimentally measured multivariate data based on UV-absorption high performance liquid chromatography in normal phase ((NP)HPLC-DAD) fingerprints of the oil fraction of avocado fruits, jointly with deep learning chemometric methods, have been successfully apply with the objective of differentiating and classifying a higher variability of avocado samples. Two different authentication goals have been pursued attending to: (i) the geographical origin, and (ii) the cultivar. Each classification method has been externally validated by establishing both the success/error contingency and proper quality performance metrics such as sensitivity, specificity, precision, or efficiency among others.

Section snippets

Chemicals

All solvents used were HPLC grade. N-hexane was supplied by Honeywell (Charlotte, NC) and isopropanol was provided by Panreac Química (Barcelona, Spain). Vitamin D3 99% supplied by Alfa Aesar (Haverhill, MA) was used as control standard.

Samples and sample preparation

One of the biggest challenges for the development of non-targeted chromatographic methods aimed at food authentication is to become available representative samples of geographic origin, plant cultivar, season, processing conditions and other factors under

Results and discussion

A two-way data array was acquired and recorded for each avocado sample. As it was stated above, the chromatogram profile at 220 ​nm for each sample was selected as characteristic fingerprint to build the different multivariate methods. Fig. 1a illustrates the heat map of a Spanish avocado sample from Hass cultivar; Fig. 1b shows the superposed chromatographic profile at 220 ​nm of the 108 avocado samples.

Conclusion

In this study, a methodology for distinguishing avocado fruits using their entire lipid chromatographic fingerprint jointly with deep learning chemometric methods is described. Different multivariate classification strategies have been tested and the differentiation of avocados according to their geographical origin and their cultivar has been achieved. To assess the different classification scenarios and describe the performance of the applied classification methods, several quality

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

CRediT authorship contribution statement

Sandra Martín-Torres: Conceptualization, Methodology, Validation, Formal analysis, Investigation, Writing - original draft. Ana M. Jiménez-Carvelo: Methodology, Validation, Formal analysis, Writing - review & editing. Antonio González-Casado: Investigation, Resources, Writing - review & editing. Luis Cuadros-Rodríguez: Conceptualization, Resources, Writing - review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements

We are grateful to Salvador Salazar for freely providing a significant number of samples from his avocado plantation located at Salobreña (Granada, Spain) to be included in this study. In addition, one of the authors, S.M.T. wants to express their sincere gratitude to the Spanish Ministry of Sciences. Innovation and Universities for a pre-doctoral fellowship (project RTC-2017-6170-2).

References (24)

  • European Union

    Regulation (EC) No 387/2005 Amending Regulation (EC) No 831/97 Laying Down Marketing Standards Applicable to Avocados

    (2005)
  • M. Flores et al.

    Avocado oil: characteristics, properties, and applications

    Molecules

    (2019)
  • Cited by (10)

    • Multi-element and stable isotopes characterization of commercial avocado fruit (Persea americana Mill) with origin authentication purposes

      2022, Food Control
      Citation Excerpt :

      These results were based on unsupervised analysis (PCA), and no classification rates were available. Martín-Torres et al., 2020 and Jiménez-Carvelo et al., 2021 combined lipid fingerprints acquired by HPLC-DAD and GC-FID, with chemometrics to successfully authenticate the origin and the botanical variety of avocados. The authors were able to discriminate avocados produced in Spain from those from other countries, including Peru, Mexico, South Africa and Kenya.

    • Combining stable C, N, O, H, Sr isotope and multi-element with chemometrics for identifying the geographical origins and farming patterns of Huangjing herb

      2021, Journal of Food Composition and Analysis
      Citation Excerpt :

      To avoid the threat of fraudulent products with a different origin, it is necessary to accurately identify the geographical origins of PGI Huangjing products and recognize the green and harmless or organic products. Many analytical approaches, based on near-infrared (NIR) spectroscopy (Chen et al., 2020), nuclear magnetic resonance spectroscopy (Cao et al., 2020), ultra-performance liquid chromatography coupled with high-resolution mass spectrometry (UPLC-HRMS) (De Carvalho et al., 2019), chromatography fingerprinting (Martín-Torres et al., 2020; Vera et al., 2019; Esteki et al., 2017), inductively coupled plasma mass spectrometry analysis (ICP-MS) (Baroni et al., 2015), and E-nose, E-tongue, and E-eye (Xu et al., 2019) have been developed to accurately identify the geographical origins and verify the authenticity of Huangjing. However, the methods mentioned above for traceability or authentication have all been based on chemical composition analysis, which varies with intrinsic genetic and metabolic factors to a large extent rather than with its origin.

    • HPLC-DAD fingerprints combined with chemometric techniques for the authentication of plucking seasons of Laoshan green tea

      2021, Food Chemistry
      Citation Excerpt :

      However, complete chromatographic separation, a necessary requirement during HPLC-UV analysis, is difficult to achieve for complex sample matrix, resulting in a decrease of fingerprint information and an increase of analytical errors of fingerprint data. In addition, only single-wavelength signals used in HPLC-UV fingerprints will also led to a low utilization of fingerprint information (Martín-Torres, Jimenez-Carvelo, Gonzalez-Casado, & Cuadros-Rodriguez, 2020). Fortunately, chemometricians have developed diverse chemometric techniques for deconvolution and resolution of spectral and chromatographic data (G.M. Escandar, Goicoechea, Munoz de la Pena, & Olivieri, 2014; Graciela M. Escandar & Olivieri, 2019).

    • Chromatographic methods

      2020, Food Authentication and Traceability
    View all citing articles on Scopus
    View full text