Evaluation of NMR-based strategies to differentiate fresh from frozen-thawed fish supported by multivariate data analysis

This dataset is the basis for a publication regarding the differentiation of fresh and frozen-thawed fish via nuclear magnetic resonance spectroscopy (https://doi.org/10.1007/s00217-023-04383-x).

The differentiation of fresh and frozen-thawed fish is a relevant authenticity aspect as in the European Union fish holds a high statistical risk of being adulterated. Here, nuclear magnetic resonance spectroscopy (NMR) in combination with principal components analysis followed by linear discriminant analysis (PCA-LDA) was used for a non-targeted based differentiation of fresh from frozen-thawed fish. To identify the most promising NMR approach(es), six different approaches were applied to 96 fish samples (mackerel, trout, cod). These approaches included different sample preparation procedures and different NMR methods to investigate both the lipid fraction and the polar fraction of the fish samples. After cross-validation embedded in a Monte Carlo resampling design, six independent classification models were obtained. Evaluation of the multivariate data analysis revealed that the most promising approaches were the 1H NMR analysis of the lipid fraction (correct prediction of about 90.0%) and the 1H NMR based screening of minor components of the lipid fraction with a correct prediction of about 91.9%. 1H NMR analysis of the water extract of the fish samples showed a correct prediction of about 82.6%.

The datasets for the six different approaches are uploaded as MATLAB workspaces, respectively. They include the NMR spectra (SPEC), SampleIDs, (fat) sample weights, internal standard weights, and the underlying groups (frozen-thawed or fresh). A guide for the SampleID is also included.