Accurate predictions of cellular response using QSPR: a feasibility test of rational design of polymeric biomaterials

We present a Surrogate (semi-empirical) model for prediction of cellular response to the surfaces of biodegradable polymers that have been designed for tissue engineering applications. The predictions of our model, when tested against experimental results, show a high degree of accuracy that is sufficient for rational design of polymeric materials for biomedical applications. The model was determined by fitting experimental data for a series of 62 polyarylates to a small number of polymer structure-based ‘molecular descriptors’ using the technique of partial least squares (PLS) regression. While PLS is commonly applied in quantitative structure activity relationship (QSAR) analysis employed in the pharmaceutical industry, this study marks the first time the technique has been extended to the problem of biomaterials discovery/design. Quantitative predictions of cellular response to six polymers (untested prior to model building) concurred with experiment within 15.8% on average. This performance compares quite favorably with the overall variation in experimental values for the library of polyarylates. Examination of the PLS ‘loadings’ reveals those structure-based features most associated with variations in the polymer performance properties, thereby providing direct guidance to the synthetic chemist in biomaterials design.