Optimal structural characteristics of osteoinductivity in bioceramics derived from a novel high-throughput screening plus machine learning approach.

Abstract

Osteoinduction is an important feature of the next generation of bone repair materials. But the key structural factors and parameters of osteoinductive scaffolds are not yet clarified. This study leverages the efficiency of high-throughput screening in identifying key structural factors, performs screening of 24 different porous structures using 3D printed calcium phosphate (CaP) ceramic scaffolds. Based on in vitro and in vivo evaluations, along with machine learning and nonlinear fitting, it explores the complex relationship between osteoinductive properties and scaffold configurations. Results indicate that bone regenerative ability is largely affected by porosity and specific surface area (SSA), while pore geometry has a negligible effect. The optimal structural parameters were identified as a porous structure with SSA of 10.49-10.69 mm² mm^-3 and permeability of 3.74 × 10^-9 m², which enhances osteoinductivity and scaffold properties, corresponding to approximately 65 %-70 % porosity. Moreover, nonlinear fitting reveals specific correlations among SSA, permeability and osteogenic gene expressions. We established a data-driven high-throughput screening methodology and proposed a parametric benchmark for osteoinductive structures, providing critical insights for the design of future osteoinductive scaffolds.