iProbiotics

iProbiotics predicts probiotic properties from whole-genome primary sequences using k-mer compositional features and machine learning to identify genomic determinants of probiotic function.

Key Features:

• Comprehensive dataset utilization: Uses genomic data compiled from the Probiotic Database (PROBIO) and extensive literature surveys.
• K-mer compositional analysis: Computes k-mer frequencies on strain genomes for k = 2–8 nucleotides to identify oligonucleotide compositions enriched in probiotic genomes.
• Incremental Feature Selection (IFS): Refines an initial set of 87,376 k-mers using IFS to isolate 184 core features that maximize model performance.
• High prediction accuracy: Reports predictive performance with accuracy 97.77% and area under the curve (AUC) 98.00%.
• Functional genomic analysis: Integrates annotations from Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), and RAST and examines genes linked to host gastrointestinal survival/settlement, carbohydrate utilization, drug resistance, and virulence factors.
• Insight into probiotic mechanisms: Indicates probiotic function is associated with combinations of k-mer genomic components rather than single-gene determinants.

Scientific Applications:

• Probiotic strain identification: Identifies candidate probiotic strains, including lactic acid bacteria consortia commonly found in food, from whole-genome sequences.
• Experimental prioritization: Prioritizes strains for experimental validation by predicting probiotic properties from genomic features.
• Functional interpretation: Links predictive k-mer features to GO, KEGG, and RAST annotations and to genes involved in gastrointestinal survival, carbohydrate utilization, drug resistance, and virulence.
• Microbiome research and probiotic development: Supports genomic analyses that inform microbiome studies and probiotic development efforts via high-accuracy predictions.

Methodology:

Selects a comprehensive dataset (PROBIO and literature), performs k-mer analysis (k = 2–8), applies Incremental Feature Selection to reduce 87,376 k-mers to 184 core features, and integrates GO, KEGG, and RAST annotations for functional analysis.