Diet and Genotype Shape the Intestinal Microbiota of European Sea Bass (Dicentrarchus labrax): Insights from Long-Term In Vivo Trials and Machine Learning

To reduce dependence on oceanic resources, poultry-based ingredients and fortified feeds have become valid alternatives to fish meal (FM) and fish oil (FO). While their impact on growth performance is well established, effects on host-associated microbiota remain less characterized. This study examines the gut microbiota of European sea bass (Dicentrarchus labrax) following FM and FO replacement with poultry- and additive-based diets, applying machine learning (ML) to evaluate diet and genotype effects. A secondary analysis of microbial profiles from two prior trials employed classification models to determine associations between microbial abundance and categorical groupings, and regression models to assess the predictive power of ingredient variations on microbial abundance. Regressors showed limited predictive capacity, whereas classifiers performed better, particularly when genotype was considered. For poultry-based diets, average accuracy was approximately 0.4 for synergistic effects, 0.6 for diet effects, and 0.8 for genotype effects; for fortified-feed diets, average accuracy was approximately 0.2, 0.4, and 0.5, respectively. Feature selection detected microbial genera encompassing beneficial (Brevundimondas, Clostridium, Idiomarina, Lactobacillus, Marinobacter, Pseudoalteromonas, Salinisphaera), neutral (Enterovibrio, Flavobacterium, Photobacterium), opportunistic (Acinetobacter, Escherichia-Shigella, Streptococcus), and undercharacterized (Acholeplasma, Cutibacterium, Enhydrobacter, Micrococcus, Peptoniphilus, Salegentibacter) taxa. ML techniques thus reveal diet- and genotype-specific signatures, underlining the importance of integrated computational-microbiological pipelines.