Associate Professor Rutgers University, United States
Introduction: Alginate is a biomaterial that has found numerous applications in biomedical science and engineering due to its favorable properties, including biocompatibility and ease of gelation. Alginate hydrogels have been particularly attractive in wound healing, drug delivery, and tissue engineering, as these biocompatible gels retain structural integrity and can be manipulated to play several critical roles. Along with its wide array of uses, alginate gels have a wide array of different formulations, each developed with its specific use case in mind. In brief, alginate hydrogel formulations can be varied in the concentration of alginate polymer used, type of crosslinker and the concentration of crosslinker. This study aims to synthesize alginate hydrogels of various formulations and study the ability to deliver a protein payload. The release kinetics of the protein release will be modeled using machine learning to map out the effects of the various formulation changes. Results from this study as a useful platform for other groups looking to predict the release kinetics of an alginate hydrogel prior to experimental testing.
Materials and
Methods: Materials: Purified sodium alginate from brown seaweed of different molecular weights, G/M ratios and concentrations were used for crosslinking. The hydrogels were crosslinked using various crosslinkers including calcium chloride, calcium sulfate, calcium carbonate, barium chloride, zinc chloride and strontium chloride. All hydrogel synthesis was performed in pH 7.4 phosphate buffered saline (PBS). Bovine serum albumin (BSA) was used as a model protein payload. Alginate Synthesis: Alginate hydrogels were prepared in 96-well plates by adding the sodium alginate, crosslinker and protein payload in a 2:1:1 volumetric ratio. Sodium alginate of different concentration was prepared by dissolving into PBS. Next, a fixed 1% BSA in PBS solution was prepared and added to the well and mixed with the alginate solution. Lastly, crosslinkers of various types and concentrations were added to the wells and mixed via shaking. The alginate hydrogels were allowed to cure for 2 hours at 37°C to ensure complete gel formation. Drug Release Study and Modeling: After successful hydrogel formation, payload release was measured by adding and aspirating PBS at set timepoints between 0h – 72h. The protein release was quantified via BCA protein quantification assay. The release profiles of the various formulations were plotted and subsequently modeled in Python using the Korsmeyer-Peppas model. The protein release of the hydrogels was modelled using a random forest regressor using the scikit-learn python package.
Results, Conclusions, and Discussions: By testing a wide array of alginate hydrogel formulations, the dataset was sufficient to train a machine learning model for produce accurate predictions of protein release curves at specified alginate formulations (Figure 1a.). Additionally, the predicted protein release curves were fitted using the Korsmeyer-Peppas model. The model constants were recorded and found to be similar to the experimental values (Figure 1b.). These results showcase the accuracy of the developed machine learning model. A comparison between predicted and actual release profiles showed more accurate predictions for lower alginate concentrations and less accurate predictions at higher alginate concentrations. This may be due to the higher alginate concentrations to produce a larger burst release profile and thus having less consistent trends which will lead to more challenging predictions for the machine learning model. In conclusion, this preliminary exploration showcases the potential for utilizing machine learning for alginate hydrogel formulations. This study looks to serve as a basis and starting point for future studies involving the use of alginate hydrogels and may serve as a reliable resource when deciding on the hydrogel formulations. Future directions for this study will look to broaden the dataset via the inclusion of additives such as polyvinyl alcohol (a known additive to influence alginate hydrogel pore sizes). We also aim to introduce additional modeling and analysis methods such as the shapely additive explanations (SHAP) to obtain a more in-depth insight into the most impactful contributors to the drug release profile.
