Many drug delivery systems have been developed to provide sustained release of proteins in vivo. However, the ability to predict and control the rate of release from delivery systems is still a challenge. Toward this goal, we screened a random drug-binding peptide library (12 amino acids) to identify peptides of varying (i.e. low, moderate, and high) affinity for a model polysaccharide drug (heparin). Peptide domains of varying affinity for heparin identified from the library were synthesized using standard solid phase chemistry. A mathematical model of drug release from a biomaterial scaffold containing drug-binding peptide domains identified from the library was developed. This model describes the binding kinetics of drugs to the peptides, the diffusion of free drug, and the kinetics of enzymatic matrix degradation. The effect of the ratio of binding sites to drug, the effect of varying the binding kinetics and the rate of enzymatic matrix degradation on the rate of drug release was examined. The in vitro release of the model drug from scaffold containing the peptide drug-binding domains was measured. The ability of this system to deliver and modulate the biological activity of protein drugs was also assessed using nerve growth factor (NGF) in a chick dorsal root ganglia (DRG) neurite extension model. These studies demonstrate that our rational approach to drug delivery system design can be used to control drug release from tissue-engineered scaffolds and may be useful for promoting tissue regeneration in vivo.