Valenzuela, Loreto M.. Experimental study and computer modeling of hydration-related behavior of L-tyrosine-derived polyarylates. Retrieved from https://doi.org/doi:10.7282/T3G44QG9
DescriptionWater uptake influences many properties of polymers and has been widely studied. In the context of polymeric biomaterials, several recent publications reported an unusual high variability of analytical results of water uptake. In the current investigation, two possible causes for the high variability of water uptake data are studied: (1) variations in the initial molecular weight of the polymer samples and (2) variations in the processing conditions used during sample preparation. Using model polymers from the combinatorial library of L-tyrosine-derived polyarylates, it was shown that the water uptake variability could be reduced significantly by annealing the film specimens after pressing. With the introduction of an annealing step, accurate and reproducible results (relative SD < 11 %) could be obtained using a 3H-radiolabeled water method that enables parallel measurements required for the efficient screening of multiple polymer samples.
Water uptake from a subset of 23 polymers from this library could not be predicted using a single parameter such as glass transition temperature or hydrophobicity. Thus, a semi-empirical model using artificial neural networks was developed to predict with high accuracy (Pearson coefficient > 0.6) the water uptake, represented by the Weibull equation. Accurate predictions (within experimental error) of water uptake were obtained for 10 of the 18 polymers used in this study, with only one polymer for which predictions were very inaccurate. The model was evaluated in an external polymer set and showed high accuracy. A semi-empirical model was also obtained for degradation kinetic parameters, with accurate predictions (Pearson coefficient = 0.7) for the kinetic coefficient of the first order model, suggesting a first order mechanism.
Predictions of water uptake and degradation kinetics were obtained for the rest of the library. These predictions may be used to select a group of polymers that satisfy certain design criteria, and eventually find a lead polymer for a specific medical application. However, modeling does not eliminate the need to run experiments, it only reduces the space of polymers that should be tested to find that lead polymer.