Towards a multi-species prediction model of fillet fatty acid composition in marine farmed fish

Abstract

Fish meal and fish oil are finite natural resources, and their use in aquaculture industry has been progressively reduced. Vegetable ingredients are the most obvious alternative, but vegetable oils are devoid of n-3 long-chain polyunsaturated fatty acids (n-3 LC-PUFA), and fish fed these oils have a fillet reduction of eicosapentaenoic (EPA, 20:5n-3) and docosahexaenoic (DHA, 22:6n-3) fatty acids. Therefore, holistic approaches for multispecies predictive modelling of fillet fatty acid (FA) composition are of relevance to guarantee the human healthful benefits of farmed fish consumption. The robustness of dummy regression analysis for predictive modelling of fillet FA composition was proven in turbot and sole using gilthead sea bream as a reference subgroup species (Ballester-Lozano et al., 2014). The absence of a statistically significant interaction between fish species and diet composition spares the necessity of the use of a vast array of diets for all the species included in the model. However, most outputs are species-specific, and each new species in the model should be introduced as a dummy regression variable and the resulting FA descriptors validated thereafter with additional data. This interactive procedure is carried out herein with European sea bass (Dicentrarchus labrax), using fish reared at the laboratory scale for the definition of the specific FA descriptors and farmed fish for the up-scaling validation of the predictive results of fillet FA composition.