Identifying factors controlling protein release from combinatorial biomaterial libraries via hybrid data mining methods

Polyanhydrides are a class of degradable biomaterials that have shown much promise for applications in drug and vaccine delivery. Their properties can be tailored for controlled drug release, drug/protein stability, and immune regulation (adjuvant effect). Identifying the relationship between the molecular structures of the polymers and the drug release kinetics profiles would help understand the release mechanism and aid in the accurate prediction of drug release and the rational design of polymer-based drug carrier systems. The molecular structure descriptors that had the most impact on the release kinetics were identified using a prediction/optimization data mining approach. Using this new approach for modeling nonlinear release kinetics behavior, we determined that the descriptors which had the greatest effect on the release kinetics were the number of backbone -COO- nonconjugated bonds, the number of aromatic rings, and the number of -CH ₂- bonds.