Model-informed prediction of antiviral drug combination synergy from sparse data using complementary mechanistic and pharmacodynamic approaches.

Introduction: Optimization of antiviral drug combinations remains challenging because exhaustive measurement of dose-response spaces is experimentally impractical. Although Bliss-based pharmacodynamic models are commonly used to describe combination responses, their empirical nature limits biological interpretation. Methods: To address these limitations, we developed a model-informed framework to reconstruct antiviral combination response surfaces from sparse diagonal observations by combining a Bliss-based pharmacodynamic model with a mechanistic viral dynamics model. Results: Both models inferred unmeasured response regions and quantified synergy relative to Bliss independence through an interaction parameter. The mechanistic viral dynamics model reconstructed response surfaces more accurately than the Bliss-based model while retaining a biologically interpretable structure linked to viral infectivity and production. Conclusion: These results provide a practical modeling strategy for antiviral combination optimization under limited-data experimental settings.