Article Preview
TopIntroduction
Identification and quantification of fatty acids is a typical analytical procedure carried out over various agricultural and food matrices. The information obtained from these analyses can be used to access a number of intrinsic characteristics of the lipid portion in these products, such as chemical quality, healthiness, stability, origin, authentication and adulteration (Laroussi-Mezghani et al., 2015; Yun & Surh 2012; Yang et al., 2013).
In order to indentify and quantify fatty acids in food, there are some standardized methods, from which the more common are those reported by The American Oil Chemists’ Society (AOCS), International Standard Organization (ISO) and International Union of Pure and Applied Chemistry (IUPAC). In general, these methods are based on preparation of methyl esters from fatty acids by esterification in alkaline or acid conditions followed by separation using gas chromatography (GC) with flame ionization detection. A crucial step of this analytical procedure is the identification of the fatty acids, which is commonly performed by comparison of the retention times of the fatty acids methyl esters (FAME) in the sample with the FAME in a standard mixture. This FAME standard mixture can be either prepared in laboratory or can be commercially available. However, the possibility of identification of a FAME in a sample is restricted by the FAME present in the standard mixture. For instance, if a chromatographic peak of a given sample does not correspond to any signal of the standard FAME mixture, its identification would not be possible, unless FAME candidates are analyzed using trial and error or the retention time of the possible FAME is somehow found/predicted.
Accordingly, some studies report the prediction of chromatographic retention times using Quantitative Structure-Property Relationship (QSPR) approaches. QSPR was employed to model retention times of 368 pesticide residues in animal tissues separated by GC using number of nitrogen atoms, solvation connectivity index‐Chi 1, Balaban Y index, Moran autocorrelation‐lag 2/weighted by atomic Sanderson electronegativity, total absolute charge and radial distribution function-6.0/unweighted as descriptors (Dashtbozorgi et al., 2013). Similarly, first component WHIM index (E1v), highest eigenvalue n.7 of burden matrix/weighted by atomic van der waals volume (BEHv7), average connectivity index Chi-2 (X2a), 3D-MoRSE signal 23 weighted by atomic Sanderson electronegativity (MoR23m) and principal moments of inertia B (PMIB) were also used as descriptors to predict the retention times of pesticides in a QSPR study (Hadjmohammadi et al., 2007). The retention times of peptides separated by liquid chromatography (LC) were predicted by QSPR based on several classes of theoretical descriptors (Golmohammadi et al., 2015). QSPR models able to predict retention times for volatile organic compounds analyzed by GC (Sarkhosh et al., 2012) and for herbicides eluted by LC (Torrens & Castellano 2012) have been reported. The results of these studies revealed the reliability and good predictability of QSPR models to predict the retention times of molecules eluted by GC or LC. In some cases, chromatographic retention times of e.g. ethers (Golmohammadi et al., 2015), peptides (Jiao et al., 2014), and even general organic compounds (Luan et al., 2008) have been shown to be better described using non-linear techniques such as Support Vector Machines (SVM) or Artificial Neural Network (ANN).