Physics-Informed Materials Informatics for Protein Adsorption on Nanostructured Titanium Oxide: Structure–Property Relationships and Inverse Surface Design

Protein adsorption on titanium-based surfaces is governed by interacting material, surface, and protein properties that complicate quantitative prediction across different experimental systems. This study presents a physics-informed materials-informatics framework for investigating protein adsorption on titanium and titanium oxide surfaces by combining a literature-derived experimental database with computational modeling and machine-learning methods. The empirical database comprises 354 experimental observations extracted from published studies. A separate computational design space comprising 689,040 physically admissible states across 11 protein systems was constructed to examine model behavior across combinations of protein and surface parameters. Computational performance was evaluated against assigned reference targets using the coefficient of determination (R2) and root mean square error (RMSE) for equilibrium surface protein density and fluorescence intensity. Across the 11 protein systems, density (R2) values ranged from −80.8615 to 0.99892, with RMSE values ranging from 0.91 to 130.44 ng/cm2. For fluorescence intensity, (R2) ranged from −4.5118 to 0.9978, with RMSE values ranging from 0.12 to 18.93 arbitrary units. These results demonstrate substantial variation in agreement with computational reference targets across protein systems, including negative (R2) values for some systems. Because the metrics were calculated exclusively against computationally assigned targets, they do not establish independent experimental predictive accuracy or empirical generalization to unseen proteins. Comparisons with external literature data provide a separate, qualified assessment of consistency with reported adsorption behavior. Overall, the framework integrates literature-derived evidence, physics-informed assumptions, and computational exploration to characterize protein adsorption across a broad parameter space. Independent validation against matched experimental measurements remains necessary to establish quantitative predictive performance.