Integrating combinatorial and computational biology for scalable ligand discovery

Abstract

Protein ligands can alter protein structure, activity, localization, and signaling, producing profound biological effects. Accordingly, the discovery and characterization of protein ligands are central challenges in biological research and medicine. Many common ligands—such as peptides and antibodies—are completely proteinaceous, consisting of linear combinations of the 20 canonical amino acids. However, because the number of unique combinations scales exponentially with protein length, sequence space is prohibitively large even for short biologically relevant sequences. This dissertation develops experimental and computational strategies for efficiently searching peptide and antibody sequence space to identify ligands for protein targets. In Chapter 2, we describe the development of a very large combinatorial peptide library displayed on the surface of yeast. This library comprises more than 6 × 109 unique peptides and covers ~23% of all possible peptides eight amino acids or shorter. We validated this library through selection campaigns against two commonly used research antibodies, an E3 ubiquitin ligase, and a bacterial enzyme complex involved in peptidoglycan synthesis. In each case, we identified motifs associated with binding and discovered multiple high-affinity, target-specific peptides. Chapter 3 translates this platform into practical, step-by-step protocols, including guidance on target preparation, selection design, sequencing analysis, troubleshooting, and interpretation. To further facilitate adoption, the library has been made commercially available at low cost. In addition to peptide discovery, this dissertation presents new screening strategies for single-domain antibody fragments, or nanobodies. In Chapter 4, we describe a strategy for identifying bacteriolytic nanobodies that combines yeast surface display techniques with a phenotypic plasmid release screen for lysis activity. Using this approach, we identified functional nanobody inhibitors of LolA, an essential bacterial protein. Finally, in Chapter 5, we present a fully computational nanobody discovery strategy based on a virtual combinatorial library. We show that AlphaFold-Multimer can accurately distinguish nanobodies that bind G protein coupled receptors (GPCRs) from non-binding clones, and we leverage this observation to perform a prospective virtual screen against the biochemically intractable itch receptor MRGPRX2. This screen yielded three functional receptor antagonists, whose binding and activity we validated with biochemical and cellular assays. Together, these studies establish scalable combinatorial and computational screening strategies to discover peptide and nanobody ligands against challenging protein targets.