Investigating the Effect of Processing and Material Parameters of Alginate Dialdehyde-Gelatin (ADA-GEL)-Based Hydrogels on Stiffness by XGB Machine Learning Model

Abstract

To address the limitations of alginate and gelatin as separate hydrogels, partially oxidized alginate, alginate dialdehyde (ADA), is usually combined with gelatin to prepare ADA-GEL hydrogels. These hydrogels offer tunable properties, controllable degradation, and suitable stiffness for 3D bioprinting and tissue engineering applications. Several processing variables affect the final properties of the hydrogel, including degree of oxidation, gelatin content and type of crosslinking agent. In addition, in 3D-printed structures, pore size and the possible addition of a filler to make a hydrogel composite also affect the final physical and biological properties. This study utilized datasets from 13 research papers, encompassing 33 unique combinations of ADA concentration, gelatin concentration, CaCl2 and microbial transglutaminase (mTG) concentrations (as crosslinkers), pore size, bioactive glass (BG) filler content, and one identified target property of the hydrogels, stiffness, utilizing the Extreme Boost (XGB) machine learning algorithm to create a predictive model for understanding the combined influence of these parameters on hydrogel stiffness. The stiffness of ADA-GEL hydrogels is notably affected by the ADA to GEL ratio, and higher gelatin content for different ADA gel concentrations weakens the scaffold, likely due to the presence of unbound gelatin. Pore size and the inclusion of a BG particulate filler also have a significant impact on stiffness; smaller pore sizes and higher BG content lead to increased stiffness. The optimization of ADA-GEL composition and the inclusion of BG fillers are key determinants to tailor the stiffness of these 3D printed hydrogels, as found by the analysis of the available data.

Keywords

• machine learning
• ADA-GEL
• osteochondral regeneration
• XGBoost algorithm