UCLA

Many actively pursued pharmacological targets are difficult to drug using conventional small molecule therapeutics because they lack conventional binding sites. These so called ‘undruggable’ targets typically interact with other proteins via shallow, solvent exposed interfaces. Peptidomimetic macrocycles have the potential to mediate such systems because the embedded peptide can mimic native protein structure and recognition elements, and the ring structures contribute to structural preorganization, lowering entropic penalties upon target binding. Compounds of this type having precise shapes and drug-like character are coveted, but are relatively difficult to synthesize. Our lab has developed methods to synthesize shape-defined macrocycles from small linear peptides. These experiments run as processes, wherein designed templates react incrementally with unprotected oligomers to form composite products. The resulting compounds retain molecular recognition elements in the oligomer, yet display that functionality as part of stable polycyclic structures. Our experimental work is based on proteinogenic amino acids and the reactivity of their nucleophilic side chains. However, using unnatural amino acids, the hypothetical scope of the chemistry becomes vast and far outpaces the capacity of our experimental format. Here, we describe the development of a computational rendering of our experimental platform, Composite Peptide Macrocycle Generator (CPMG). This open-source platform simulates our multi-step reaction chemistry using a large, tailored monomer set. We have used the algorithms to anticipate product outcomes of >2 billion processing sequences. We have further developed software to generate three- dimensional structures for each product. Every library member has feature constraints meant to increase the probability of it being bioavailable. We discuss efforts to merge our experimental and computational abilities into a single, iterative workflow to discover new macrocyclic ligands for challenging protein targets. We describe new computational tools and techniques to allow rapid, flexible docking of conformationally dynamic ligands onto multiple protein targets. We describe experiments to validate predictions by synthesizing novel arene amino acids and engaging them in macrocyclizations to generate previously unknown ring systems. We show how the interplay between calculations and synthesis can offer insight into molecular properties and applications.