Genetic and phenotypic validation of Raman-predicted fatty acids in Atlantic Salmon: From grouped fatty acids to EPA and DHA

Summary

This study evaluates the potential of Raman spectroscopy as a high-throughput phenotyping tool for predicting fatty acid (FA) composition in Atlantic salmon (Salmo salar) fillet muscle. A total of 33 FA traits, including individual and grouped saturated (SFA), monounsaturated (MUFA), and polyunsaturated fatty acids (PUFA), were analyzed from 613 samples using gas chromatography (GC) as the reference method. Partial least squares regression (PLSR) combined with Markov Blanket (MB) feature selection was applied to develop Raman-based prediction models on a subset of 100 samples. Genetic parameters were estimated to assess the relationship between directly measured FA traits and Raman-predicted values on 506 independent samples as a genetic validation. This study found that 17 of 33 traits exhibited strong genetic correlations (RG > 0.85) with directly measured GC values, supporting their potential for genetic selection. Individual FA prediction was more challenging for traits with low abundance or low heritability estimates of a given FA trait. In contrast, Raman-predicted eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA) showed high genetic correlations with reference measurements (RG = 0.88), with a combined EPA + DHA trait achieving an even stronger correlation (RG = 0.97). Overall, these results demonstrate that Raman spectroscopy holds significant promise for large-scale phenotyping and genotyping of key FAs such as EPA and DHA in Atlantic salmon breeding programs. Future research should focus on improving data acquisition methods, including rapid full-fillet scanning, robotic integration, and through-skin measurements as potential approaches, along with advanced modeling strategies to maximize utility under commercial conditions.