Prediction of naturally-occurring, industrially-induced and total trans fatty acids in butter, dairy spreads and Cheddar cheese using vibrational spectroscopy and multivariate data analysis

Abstract

This study investigated the use of vibrational spectroscopy (i.e. near infrared (NIR), Fouriertransform mid infrared (FT-MIR), Raman) and multivariate data analysis for (1) quantifying total trans fatty acids (TT) and (2) separately predicting naturally-occurring (NT; i.e. C16:1 t9; C18:1 trans-n, n=6…9, 10, 11; C18:2 trans) and industrially-induced TFAs (IT=TT-NT) in Irish dairy products i.e. butter (n=60), Cheddar cheese (n=44), dairy spreads (n=54).Results of gas-chromatography (GC) analysis were used as the chemical reference values.Butter, dairy spreads and Cheddar cheese separately contained TT around 4.42%, 0.70% and 3.04% of their total fatty acid methyl ester (%w/w FAME). Butter contained about 4.22% Prediction of NT, IT and TT in Cheddar cheese was with low accuracy due to interference from protein; in the case of dairy spreads and butter, prediction of IT was inaccurate due to the limitation of vibrational spectroscopy in predicting very low analyte contents (< 0.2% w/w of TFA).

Invited review: Use of infrared technologies for the assessment of dairy products—Applications and perspectives

“…Cheddar cheese from one dairy factory predicted more accurately the cheese composition in other 3 factories than NIR spectroscopy calibrated in each factory. On the other hand, total trans FA and natural occurring trans FA in butter were better predicted with NIR than MIR spectroscopy (Zhao et al, 2015). A similar probability of correct classification for Parmigiano Reggiano authentication has been reported for Fourier-transform NIR and MIR (Cevoli et al, 2013).…”

Investigation of Raman Spectroscopy (with Fiber Optic Probe) and Chemometric Data Analysis for the Determination of Mineral Content in Aqueous Infant Formula

“…645, 763 and 877 cm −1 are also relevant to δ (C-C-O) or δ (C-C-H) bonds of tryptophan [21,46]; 950 cm −1 has been assigned to δ (C-O-C), δ (C-O-H) and ν (C-O) bonds; 1065-1082 and 1121cm −1 has also been assigned to δ (C-O-H), ν (C-O) and ν (C-C) of aspartic and glutamic acid [47]. Raman bands at 1003-1005 cm −1 have been strongly related to the ring-breathing structure of phenylalanine reported in many previous studies [21,41,47,48]. Peaks at 1262, 1303, 1442 and 1745-1748 cm −1 are assigned to γ (CH 2 ), τ (CH 2 ), δ (CH 2 ) and ν (C=O) bonds of aliphatic chains in lipids and amino acid residues, respectively [21,48].…”

“…The satisfactory prediction results from a PLSR model were expected to have R 2 values close to 1, RMSE values, and bias close to 0. Robust PLSR models were developed using a small number of latent variables (PLS loadings) [41]. Improvements in the performance of PLSR models were attempted using a reduced number of Raman spectral variables which were selected by variable importance on projection (VIP) [42], significance multivariate correlation (sMC), and Martens' uncertainty test [43].…”

Total and Free Fatty Acids Analysis in Milk and Dairy Fat

“…Zhao et al [119] investigated the use of NIR, FT-mid-infrared (FT-MIR) and Raman spectroscopies and multivariate data analysis to quantify total tFA (TT) and to detect naturally occurring (NT) and industrially induced (IT) tFA in butter, Cheddar cheese and dairy spreads. They considered NT all peaks identified by CG-FID as trans-C18:1 and trans-C18:2 isomers and t9-C16:1.…”