Molecular simulation and machine learning tools to predict bioglass modulus of elasticity

The structure-property relationship of bioactive glasses depends on costly synthesis and property tests. Quantitative models with predictive capacity, based on molecular simulation (MS), is an alternative to accelerate these developments. Here, we apply MS and machine learning (ML) to predict the modulus of elasticity and expand the test capacity on large scale, for the compositions around the bioactivity envelopes, for 45S5 bioglass (6 wt% P2O5) and also 4, 5 and 7 wt%. We validate a force field, and with a simulated database, we tested 4 different ML algorithms. Na2O has the greatest impact on the modulus of elasticity, followed by CaO. The 6% and 7% P2O5 diagrams showed more compositions with high values for the modulus of elasticity and bioactivity simultaneously. Considerations about convenience of use ML methods are also presented. This study demonstrates the potential of MS and ML techniques in decision-making for the synthesis of glasses with bioactivity.