Identifying probiotic microorganisms requires linking genomic content to experimentally reported host-interaction traits. A machine-learning framework predicted seven probiotic phenotypes (acid resistance, bile resistance, adhesion, antimicrobial activity, immunomodulation, antioxidant activity, and antiproliferative potential) from functional COG categories and biosynthetic gene cluster annotations in 1183 genomes (780 probiotic and 403 non-probiotic). CatBoost, Random Forest, XGBoost, LightGBM, and Logistic Regression were evaluated with stratified cross-validation, leave-one-out cross-validation, and an 80:20 holdout split. Antioxidant and antiproliferative models performed best, with cross-validation F1-scores of 0.92 and 0.97 and holdout F1-scores of 0.70 and 0.72. SHAP analysis identified model-derived genomic associations rather than causal mechanisms, including an inverse association between T3PKS abundance and antioxidant classification and associations between lipid transport/metabolism features and antiproliferative classification. The framework supports high-throughput in silico prioritization of candidate probiotic strains for phenotype-specific validation. Interpretation remains limited by heterogeneous source annotations, label noise, and potential false negatives in the non-probiotic dataset.
The search for next-generation probiotics derived from the human gut microbiota has intensified due to their potential beneficial effects in modulating immunity and preventing diseases. Spore-forming bacteria are attractive probiotic candidates due to their functional traits and unique resilience, which offer high stability to commercial formulations. In this study, spore-forming bacteria isolated from human fecal samples were screened. Two strains, Bacillus spizizenii A1 and Priestia megaterium F6, were identified as non-hemolytic and not resistant to a panel of antibiotics relevant for spore formers. Both strains were characterized by using in silico and in vitro approaches, demonstrating relevant probiotic potential. The comparative analysis indicated B. spizizenii A1 as the most promising probiotic candidate due to the resistance of its spores to simulated gastric conditions and for its antioxidant, antimicrobial, and antibiofilm activities. Strain A1 combines a safety profile with functional probiotic traits and is therefore a strong candidate for further in vivo evaluation and future development as a next-generation probiotic.
The identification of probiotic microorganisms is a complex process that requires inte-grating genomic data with evidence of functional health benefits. This study presents a machine learning framework, applied to a dataset of 1,185 probiotic and non-probiotic genomes, to predict seven key probiotic phenotypes (acid resistance, bile resistance, adhesion, antimicrobial activity, immunomodulation, antioxidant, and antiproliferative potential) using functional (COG) and metabolic (AntiSMASH) genomic annotations. We evaluated several algorithms, including CatBoost, Random Forest, and XGBoost, across three validation scenarios: cross-validation, leave-one-out, and train-test split. Models for antioxidant and antiproliferative activities achieved the highest predictive performance. Shapley Additive explanations (SHAP) analysis revealed specific metabolic signatures for each trait, such as the critical role of Type III polyketide synthases (T3PKS) in antioxidant activity and lipid metabolism, which contribute to antiproliferative effects. These results demonstrate that combining machine learning with functional analysis is a robust strategy for the in silico screening and selection of novel probiotic strains.