(568e) Data Mining Nature to Design New Peptide Based Biomaterials

One strategy for developing new biomaterials is creating a self-assembling peptide surface. Peptides have the flexibility of many monomer units to choose from and are easy to synthesize. However, the design space of peptides has a combinatorial explosion at any significant size. One approach to solving this problem is to use structural bioinformatics to utilize sequence patterns seen in Nature. Starting from a simple structural hypothesis, we've designed ultra-low fouling inert biomimetic surfaces from patterns observed in structural protein data. We first used protein ontologies to find relevant proteins and then used a combination of structural and sequence conservation metrics to gain an understanding of common features. This led to rationally designed peptides that can obtain ultra-low fouling when immobilized on self-assembled monolayers. One particular combination of peptides was found to be frequently occurring in Nature and repeats of that sequence created inert surfaces. Additionally, we fit Markov models to the protein ontologies and used these models to design combinatorial peptide libraries that "look like", for example, blood or the cytoplasma.