Building Rational Methods for Peptidomimetic Design: Adding Peptoids and Non-canonical Amino Acids to Rosetta
Our method for predicting protein structure, Rosetta, has been adapted to protein design. We have been adapting this code to include amino acids not found in nature. This code will be used by NYU chemist Kent Kirshenbaum and Paramjit (Bobby) Aurora to design molecules that look a lot like peptides (peptidomimetics) but are much tougher as drugs and binding agents used for imaging and diagnostics. One real problem in the field of peptidomimetics has been the lack of modeling/design tools to design these molecules in silico, this effort will provide this critical tool. Predictions made with the newest versions of our code are being OR have been tested in the Drs. Kirshenbaum and Aurora's labs resulting in two papers and a pending grant.
Our work will remove a key bottleneck in the development of protein-based diagnostics by developing methods for designing molecules that tightly and specifically bind a protein of interest. We will develop computational tools to rationally design peptide-based capture agents targeted to oncogenic protein biomarkers. Peptides are promising binding agents, as they are cheap to produce and easily permuted and optimized. However, "traditional" peptides, consisting of the 20 canonical amino acids (CAAs) found in nature, have several drawbacks as therapeutics and high affinity capture agents, despite their low cost and modular nature; peptides are subject to degradation from proteases, may have poor solubility, and attaching peptides to solid substrates often interferes with binding.
We are developing methods that can be used to design peptides containing (or entirely composed of) non-canonical amino acids (NCAAs). We will use these methods to design protease resistant NCAA-peptides that tightly bind an initial set of oncogenic proteins. Our work will build upon best-in-class platforms for protein design (we have significant experience actively developing and were initial contributors to the Rosetta code). By integrating our work with the Rosetta platform we will also gain proven ability to model/design natural proteins and peptides (as well as CAA+NCAA chimeras). We will deploy these methods in collaboration with experimental groups actively working on synthesis and combinatorial screening of NCAA-containing peptides. We know of no method for NCAA-peptide/protein design with the proposed functionality. If we are successful these NCAA-peptide high-affinity capture agents will be central to a new generation of cost-effective multi-parameter diagnostics.
In addition to optimizing affinity we will incorprate design rules and principles that will enhance protease resistance and also allow for multiple methods for attaching these molecules to solid supports (a key engineering requirement). Our methods will be built to compliment current methods for finding peptide binders, namely screening of large semi-random libraries; to this end we describe current collaborations that illustrate how our proposed method will be used to design robust biosensors/diagnostic devices.