OptiSyn: an interpretable, multi-omics-driven graph convolutional network framework for synergy-oriented drug...

Area of Science:

• Biomedical Science
• Computational Biology
• Pharmacology

Background:

• Modern biomedical science increasingly relies on bioinformatics and computational modeling for drug discovery.
• Integrating multidimensional data with systems biology and AI offers a framework for understanding Traditional Chinese Medicine (TCM).

Purpose of the Study:

• To identify key genes associated with ankylosing spondylitis (AS) using multi-omics data.
• To develop an interpretable AI model for predicting optimal drug combinations and synergistic therapeutic roles in TCM.
• To validate the efficacy of a novel TCM formula designed by the AI model.

Main Methods:

• Multi-omics datasets, differential gene expression analysis, weighted gene co-expression network analysis, single-cell transcriptomic analysis, and Mendelian randomization were used to identify AS-associated hub genes.
• An interpretable, multi-layer graph convolutional network model was constructed using network topology, molecular docking, clinical data, and compound similarity.
• In vitro and in vivo experiments were conducted to evaluate the therapeutic effects of the identified TCM formula.

Main Results:

• Eight AS-associated hub genes were identified, linked to immune responses and T-cell-mediated processes.
• The AI model prioritized a novel TCM formula, ASD-A, comprising eight herbs.
• ASD-A demonstrated significant reduction of pro-inflammatory cytokines (IL-6, TNF-α), modulated immune cell subsets, regulated key genes (KRAS, SMAD2, MAPK14), and rebalanced the IL17/Foxp3 axis in AS models.
• Ablation studies confirmed the importance of multi-source data integration for formula design.

Conclusions:

• The AI-driven approach provides novel insights into combinatorial therapy for AS, aligning with TCM principles.
• This study highlights the potential of integrating bioinformatics and AI with TCM for efficient, mechanistically informed disease treatment.
• The findings support the development of personalized therapeutic strategies by combining modern computational techniques with traditional medicine.