Machine Learning Approaches Applied to GC-FID Fatty Acid Profiles to Discriminate Wild from Farmed Salmon

Area of Science:

• Food science
• Analytical chemistry
• Chemometrics

Background:

• Increasing demand for wild-caught fish leads to higher prices and potential for mislabeling.
• Distinguishing wild from farmed salmon is crucial for consumer protection and market integrity.

Purpose of the Study:

• To develop and validate a method for discriminating between wild and farmed salmon using fatty acid profiles.
• To assess the effectiveness of various chemometric approaches, including machine learning classifiers, for salmon classification.

Main Methods:

• Lipids were extracted from salmon muscle samples (wild and farmed from various origins).
• Fatty acid composition was analyzed using gas chromatography with flame ionization detection (GC-FID).
• Chemometric techniques, including principal component analysis (PCA) and seven machine learning classifiers, were applied for data analysis.

Main Results:

• Chemometric approaches, particularly machine learning classifiers, successfully differentiated between farmed and wild salmon.
• Six out of seven classifiers achieved ≥99% accuracy, with five achieving 100% accuracy on the test dataset.
• The k-nearest neighbors (kNN) classifier achieved perfect classification using only three key fatty acid features (16:0, 18:2n6c, and 20:3n3 + 20:4n6).

Conclusions:

• Fatty acid profiling combined with chemometrics provides a robust and accurate method for distinguishing wild from farmed salmon.
• The proposed approach is simple, effective, and utilizes commonly available laboratory equipment.
• This method can aid in combating seafood mislabeling and ensuring product authenticity.