Classification of Unknown Samples by Fatty Acids

Significance of Topic

Fatty acid profiling, especially of medium-chain unsaturated fatty acids, is crucial for evaluating the nutritional and functional qualities of food by-products such as brans. Beyond simple physical characterization, multivariate classification of lipid profiles enables rapid grouping, quality control, and discovery of characteristic markers for R&D and industrial applications.

Objectives and Study Overview

This study aimed to analyze fatty acid methyl esters (FAMEs) in 48 rice bran samples, classify them into distinct clusters, identify the most discriminative fatty acids, and build a predictive model capable of assigning unknown samples to the correct group.

Methodology and Data Processing

Bran samples were derivatized and analyzed by gas chromatography–mass spectrometry. LabSolutions Insight™ software performed automatic peak integration and flagged saturated signals. A CSV file of peak areas for 37 target FAMEs was exported and preprocessed: compounds lacking integration in any sample or exhibiting detector saturation were removed. Normality tests reduced the list to 25 compounds, and pairwise correlation analysis further condensed it to 16 variables by averaging highly correlated pairs.

Used Instrumentation

• Shimadzu GCMS-QP™2020 NX
• Shimadzu GCMS-TQ™ NX series
• LabSolutions Insight™ software for data acquisition and processing
• Orange Data Mining (University of Ljubljana) for multivariate analysis

Main Results and Discussion

Principal component analysis on the 16 selected FAMEs revealed three clusters, with PC1 and PC2 explaining approximately 80 % of the variance. A classification tree identified key fatty acids—such as methyl laurate (C12:0)—as decisive in distinguishing between clusters. Validation with three previously unseen bran samples demonstrated accurate assignment to the corresponding group.

Benefits and Practical Applications

• Seamless workflow linking GC-MS acquisition to offline multivariate analysis
• Rapid identification of chemical markers for sample groups
• Reliable classification model for quality control and unknown sample prediction

Future Trends and Potential Applications

Advances may include integration with automated machine-learning pipelines, expansion to other food matrices or by-products, and development of high-throughput or real-time fatty acid profiling platforms. Deeper exploration of functional nutrient distributions could support personalized nutrition and supply-chain transparency.

Conclusion

The described GC-MS workflow, combined with LabSolutions Insight™ preprocessing and Orange Data Mining analysis, enables effective classification and prediction of bran samples based on fatty acid profiles. This approach supports R&D efforts, quality assurance, and rapid decision-making in food analysis.