Integrating molecular dynamics and polynomial regression for predicting structure–property relationships in silicate-based melt-quench derived binary bioactive glasses

Bioactive glasses (BGs) are widely utilized biomaterials in biomedical applications due to their ability to withstand high mechanical stresses and biologically active ion release behavior. In this study, the structural and mechanical properties of (100-x)SiO 2 -xCaO (x = 0, 10, 20, 30, 40, 50, 60 mol%) BGs were investigated using molecular dynamics (MD) simulations in LAMMPS software. Meanwhile, structural parameters, including radial distribution functions (RDFs), bond lengths, bond angles, coordination number (CN), bridging oxygens (BOs), non-bridging oxygen (NBOs), Q n distribution, density, and network connectivity (NC), were analyzed. Results demonstrated that the Si–O bond length remained ∼1.6 Å across compositions, while Ca–O bonds (∼2.3–2.4 Å) showed increasing CN with higher CaO content in BGs composition. The Si–O–Si angle decreased from 152.27° to 146.78° as CaO increased, while BO content dropped from 99.76 % (100Si) to 17.30 % (40Si), accompanied by a corresponding rise in NBOs and low-n Q n units, leading to reduced NC (3.990 → 1.210) and structural integrity. Mechanical properties from simulated tensile tests revealed that Young's modulus decreased from 117.12 GPa (100Si) to 67.766 GPa (50Si), yield stress from 16.592 MPa (100Si) to 6.841 MPa (40Si), while elongation peaked at 0.77 for 50Si. Additionally, fourth-order polynomial regression with Ridge regularization was employed to predict mechanical properties based on volume fraction, achieving low errors and capturing nonlinear composition–property trends. Taken together, results confirmed that the integration of MD and machine learning enabled accurate property prediction from limited datasets, facilitating the design of BGs with simultaneously optimized structural properties, and mechanical performance. The approach presented in this study provided a scalable route for engineering BGs with customized mechanical and structural characteristics, and held potential for adaptation to other BG compositions or bioceramics where an optimal combination of strength, ductility, and structural integrity is essential. The proposed approach provided a scalable route for engineering bioactive glass compositions with optimized structural integrity and elastic response at the atomic level, while acknowledging the intrinsic brittleness of bulk bioactive glasses and their primary role as components of composite or coating systems.