Modeling sRNA-mRNA regulatory interactions in Staphylococcus aureus from peri-implantitis using a graph attention autoencoder.

Abstract

Introduction: Microbial dysbiosis and chronic infection are linked to peri-implantitis, a complex inflammatory disease. Staphylococcus aureus, capable of stress adaptation, immune evasion, and antibiotic resistance, is increasingly recognized as a major cause of implant infections. These traits are regulated by small RNA (sRNA), but the regulatory environment in oral settings remains poorly understood. This study employs empirical sRNA-messenger RNA (mRNA) interaction data and a graph-based method to analyze these regulatory circuits.

Materials and methods: Using a curated cross-linking, ligation, and sequencing of hybrids dataset of sRNA-mRNA interactions in S. aureus, we implemented a hybrid graph attention autoencoder (GAT-AE). Interaction scores were encoded as edge weights to construct a bipartite graph. The model used a feedforward decoder to reconstruct interaction scores and graph attention layers to learn 64-dimensional node embeddings. Mean squared error, mean absolute error, and coefficient of determination (R ²) were used to assess the model's performance.

Results: The model identified SprX and RsaE as key regulatory sRNAs that govern antibiotic resistance, stress responses, and metabolic adaptation. Because of its role in immune evasion, the spa mRNA has become a major regulatory target. The model identified co-regulated clusters, reconstructed the overall network structure, and captured biologically significant patterns, despite modest predictive accuracy (R ² = 0.01).

Conclusion: This study demonstrates the utility of GAT-AE models for investigating posttranscriptional regulation in S. aureus, particularly in the context of peri-implantitis. Identifying sRNA hubs involved in survival and virulence lays the groundwork for future diagnostic and therapeutic approaches targeting sRNA-mediated pathways in oral infections.