Predicting inter-microbial host specificity in oral biofilms using a lightweight relation-aware knowledge graph...

Area of Science:

• Microbiology and Bioinformatics
• Computational Biology
• Systems Biology

Background:

• The human oral cavity harbors a complex microbial ecosystem, including bacteria and viruses, forming biofilms.
• Phage-bacteria specificity is vital for microbial community stability and preventing dysbiosis, but mapping it is experimentally challenging.
• Traditional methods struggle to capture the ecological complexity of oral microbiome interactions.

Purpose of the Study:

• To develop a graph-based model for predicting inter-microbial host specificity in oral biofilms.
• To integrate diverse data (taxonomic, ecological, infection) into a knowledge graph for relational learning.
• To improve predictions of phage-bacteria interactions and identify microbial hubs linked to periodontal disease.

Main Methods:

• Constructed a heterogeneous, relation-aware knowledge graph of the oral microbiome, including taxa, niches, and infection relationships.
• Integrated microbial features with graph embeddings and developed a relation-aware graph neural network (IK-BRNet).
• Employed stratified cross-validation with class imbalance correction for model evaluation against a conventional Graph Attention Network (GAT).

Main Results:

• IK-BRNet demonstrated faster convergence and superior discrimination, achieving a higher AUC-ROC (0.929 vs. 0.904) compared to GAT.
• IK-BRNet significantly improved sensitivity for disease-associated viral taxa (93.8% vs. 56.3%), reducing false negatives.
• Site-specific predictions aligned with biological validity, identifying higher disease scores for dental plaque-associated viruses.

Conclusions:

• Relation-aware graph learning provides an effective framework for modeling inter-microbial host specificity in oral biofilms.
• The developed model enhances oral microbiome network inference, aiding disease screening and ecological analysis.
• This approach supports advancements in microbiome-based dentistry and understanding oral health.