Model-driven design of bioactive glasses: from molecular dynamics through machine learning

Abstract

Research in bioactive glasses (BGs) has traditionally been performed through trial-and-error experimentation. However, several modelling techniques will accelerate the discovery of new BGs as part of the ongoing endeavour to 'decode the glass genome.' Here, we critically review recent publications applying molecular dynamics simulations, machine learning approaches, and other modelling techniques for understanding BGs. We argue that modelling should be utilised more frequently in the design of BGs to achieve properties such as high bioactivity, high fracture strength and toughness, low density, and controlled morphology. Another challenge is modelling the biological response to biomaterials, such as their ability to foster protein adsorption, cell adhesion, cell proliferation, osteogenesis, angiogenesis, and bactericidal effects. The development of databases integrated with robust computational tools will be indispensable to these efforts. Future challenges are thus envisaged in which the compositional design, synthesis, characterisation, and application of BGs can be greatly accelerated by computational modelling.

Modelling the elastic mechanical properties of bioactive glass-derived scaffolds

“…Looking at the future, further improvements in the predictive capability could be achieved by applying molecular dynamics simulations and machine learning approaches, which are already implemented for the design of bioactive glasses to tailor specific properties, such as density, dissolution kinetics and bioactivity, or model the biological response, such as the glass ability to foster protein adsorption, cell adhesion, cell proliferation and antibacterial effects [37,38]. Indeed, a critical issue to be considered concerns the incorporation of scaffold microstructure in the simulation and modelling procedures, carrying the risk to further increase the algorithmic complexity and computational cost.…”

A critical review of bioceramics for magnetic hyperthermia

“…Preceramic polycarbosilane polymers are commonly used as precursors to make ceramic fibers 10,39 and ceramic matrix. [7][8][9]40,41 The chemical and volume changes that accompany ceramization of the polymer during pyrolysis have been documented in several studies. For example, see Refs [39][40][41][42].…”

Artificial intelligence and machine learning in glass science and technology: 21 challenges for the 21st century

“…Some of the areas where reinforcement learning can have applications include the design of smart robots (robotics), which allow automated high‐throughput synthesis, characterization, selection, and design of novel materials 30 . A review of some of the algorithms used in materials science and glass science can be found elsewhere 15,34‐36 …”

“…30 A review of some of the algorithms used in materials science and glass science can be found elsewhere. 15,[34][35][36]…”