mRNA display allows production and selection of vast macrocyclic peptide libraries.  Here we present a novel strategy for making target class-selective mRNA display libraries by using N-terminal selective cyclization chemistry to allow post-translational chemical derivatization of internal cysteines. We thus install analogs of dimethyl lysine (KMe2) in selections against epigenetic targets UHRF1 and RBBP7. We further expand this methodology by combining it with a novel, late-stage barcoding strategy that allows rapid preparation of focused, bespoke mRNA display libraries for hit-to-lead optimization. This approach highlights how combined library diversification methods can enhance mRNA display for discovering potent binders to protein targets.

Phage display has become a cornerstone of peptide discovery, enabling the identification of high-affinity binders against a wide array of targets. However, binding affinity alone is not a reliable predictor of therapeutic success. Enhancing the translational potential of peptides requires addressing critical factors such as off-target effects, biodistribution, and pharmacokinetics early in the discovery process. By employing selection strategies under physiologically relevant conditions, we can prioritize candidates with optimized therapeutic profiles, leading to more effective peptide-based therapies.

Hydrolases are enzymes that often play pathogenic roles in diseases such as cancer, asthma, arthritis, atherosclerosis, and infection by pathogens. Probes that allow dynamic monitoring of their activity can be used as diagnostic and imaging agents, as well as for identification of enzymes as drug leads. I will describe efforts using phage display, mRNA display, and high-throughput fragment screening to identify selective covalent-binding probes for diverse protein targets.

We have integrated genetically encoded noncanonical amino acids (ncAAs) into the phage display technique to facilitate simultaneous cyclization of phage-displayed peptides and use the chemical handles in ncAAs as anchors for active sites of protein targets for directed ligand evolution and enrichment. To achieve the full potential of this novel technique, we have also developed chemical linkers that conjugate to an N-terminal cysteine and an internal cysteine for ligation of phage-displayed peptides. These techniques have been successfully applied for the identification of peptide ligands for epigenetic protein targets, including SIRT2, ENL, HDAC2, and BRD9. 

Affinity selection on random peptide libraries, coupled with next-generation sequencing, yields high-throughput yet sparse data, which we use to train biophysical models that predict SH2 domain binding free energy and c-Src kinase efficiency over the full theoretical sequence space. Our model predictions are validated against biophysical measurements of synthesized peptides. This unbiased approach enables scalable, accurate prediction of protein functional properties, supporting more effective identification and optimization of drug candidates.

Standard peptide discovery methods like phage and mRNA display, face issues like high false positives or costly licensing, limiting therapeutic advances. We introduce a high-throughput discovery platform merging machine learning (ML) and peptide array libraries, demonstrating high hit rates and the ability to leverage ML trained on the primary data to optimize weak binders toward nanomolar affinity. Additionally, we present visualization techniques suitable for high-throughput discovery data for binding-mode detection and offer insights into the future of ML-driven peptide optimization.

Understanding the MHC presentable peptidome is critical for rational vaccine and immunotherapy design. We developed a scalable yeast-display platform to map peptides from pathogens and tumors that are stably presented by MHC. Using HIV as a test case, we defined the viral peptidome; identified conserved, stable epitopes; and characterized potential escape mutations. This strategy provides a generalizable framework for defining peptidomes and guiding the design of T cell vaccines.

GPCRs are vital drug targets, yet remain difficult to target with biologics. We combine computational de novo design with a high-throughput, microscopy-based “receptor diversion”–pooled screen to create high-affinity, selective miniprotein agonists and antagonists. The platform produced MRGPRX1 agonists, as well as CXCR4, GLP1R, GIPR, GCGR, and CGRPR antagonists. Cryo-EM reveals atomic-level accuracy, demonstrating precise control of GPCR function and broad therapeutic potential.

Fibrosis underlies progressive dysfunction across multiple organs and remains poorly addressed by current therapies. We describe an AI-enabled peptide discovery platform informed by regenerative biology that enables rapid identification of short, multi-mechanistic peptides with anti-fibrotic activity. Lead peptides demonstrated multi-organ suppression of fibrotic and inflammatory markers in human cell systems and attenuated fibrosis in rabbit scarring and murine pulmonary fibrosis models, achieving efficacy comparable to nintedanib while preserving systemic health. These findings highlight the potential of regenerative, AI-designed peptides as a novel therapeutic approach for inflammatory and fibrotic diseases.

Topics for Discussion:

• Proteolytic stability
• Pharmacokinetics
• Oral bioavailability
• Cell surface vs intracellular targets
• Cost of goods: synthetic vs recombinant​

Topics for Discussion: 

• Strategies to improve the identification of high-affinity & selective peptide binders 
• Focus on early screening of selected hits from discovery efforts (e.g leveraging M/L models) 
• How can we prioritize candidates effectively, and what workflows can optimize this process? 
• What are the most effective methods for synthesizing peptides quickly and accurately?   
• Exploration of innovative approaches including bioproduction​

We developed a high-throughput platform for the expression and purification of peptides, miniproteins, and small-protein scaffolds, optimized for rapid target discovery and validation. Central to this workflow is an engineered SUMO protease with exceptional activity on magnetic beads, enabling efficient, contamination-free elution without buffer exchange. Integrated with automated E. coli expression, the system produces over 1000 purified proteins weekly at >95% purity and >200 µg yield per target. This 4-day DNA-to-protein pipeline accelerates peptide-screening campaigns, supports triaging of candidate molecules, and generates high-quality datasets for machine-learning-driven peptide and protein engineering.

A versatile and highly efficient bioproduction platform to generate various forms of disulfide-constrained peptides (DCPs) has been developed as an environmentally sustainable alternative to SPPS. This platform can be used to generate: (1) multivalent DCPs with different geometries, (2) DCPs with functional chemical groups such as biotin, (3) DCPs with unnatural amino acids through amber codon suppression, and (4) isotope-labeled DCPs.

We developed a semi-recombinant process for Semaglutide production using Escherichia coli. The tandem repeated precursor peptide (P29x8) is expressed at >10.0 g/L as inclusion bodies, then solubilized, refolded, and enzymatically cleaved to yield monomeric P29. Subsequent fatty-acid side-chain acylation and dipeptide ligation complete Semaglutide synthesis. The integration of high-titer fermentation with optimized enzymatic and chemical steps enables a final Semaglutide production yield exceeding 4.0 g/L.

Histatins comprise a family of ~12 histidine-rich peptides naturally present in human saliva. Their antimicrobial properties have attracted significant interest as potential therapeutics for combating oral infections. Recombinant expression of histatin peptides with E. coli has traditionally used cyanogen bromide to cleave the desired peptide sequence from a fusion protein. This talk will present an immobilized enzyme approach for obtaining histatin peptides that obviates the need for cyanogen bromide. The applicability of this approach to the production of other peptides, and its practicality in terms of yield, cost, and environmental impact, will also be discussed.

Conjugate vaccines, composed of pathogen glycans attached to immunogenic carrier proteins, are effective tools to prevent bacterial infections. However, many conjugate vaccines produce poor immune responses, and new methods are required to optimize vaccine design for stronger immunogenicity. We developed an in vitro workflow coupling cell-free gene expression and AlphaLISA to rapidly characterize and engineer post-translational modifications, including glycosylation. We used our workflow to engineer oligosaccharyltransferases involved in protein glycan coupling technology, leading to the identification of mutant enzymes and sites within a vaccine carrier protein that enable high efficiency production of glycosylated proteins. We then scaled up the cell-free production of vaccines for in vivo immune evaluation. Our method accelerates the characterization of post-translational modifications and the engineering of enzymes for more efficient production of therapeutic proteins.