Abstract
AbstractWe present two surrogate (semi‐empirical) models for prediction of both Protein Adsorption (PA) onto, and Cellular Response (CR) to, the surfaces of biodegradable polymers that have been designed for tissue engineering applications. We explore several of the key issues in the modeling process including the selection of adequate input variables, and the influence of the choice of the training set and the number of input variables on the accuracy of the predictions. The models are each built using five input variables selected from over 800 choices that include experimentally measured values, empirical estimates of chain hydrophobicity, and parameters that describe molecular shape. For both the PA and CR surrogate models, increasing the number of input variables from five up to seventeen resulted in no significant increase in model accuracy. Further, we examined the influence of the number of iterations in a Monte Carlo analysis designed to ensure that the amount of experimental error is realistically accounted for in the models. We found two ANN‐based surrogate models that predict PA and CR for 77 and 71% of the validation sets to within experimental uncertainty, respectively.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call