Info
Background:
The primary bottleneck in molecular imaging development is discovering ligands with both high affinity and disease-specific selectivity1. Disulfide-constrained peptides are particularly promising as imaging agents due to ease of manufacturing, minimal immunogenicity, and high target affinity2. Phage display offers a powerful approach for discovering peptide-based agents with high specificity; however, each target requires ligands with distinct three-dimensional architectures to properly position binding residues within the target’s binding pocket, and current phage libraries are inefficient due to their uniform structural scaffolds that require screening multiple libraries to identify optimal ligands for diverse targets3. To address this structural diversity challenge, we engineered a variable-loop-length library that exploits phage biology to sample diverse architectures in a single screen. This library, coupled with robust computational analysis, enables systematic insight into how loop size influences target recognition. By filtering out nonspecific binders and identifying structures that enable productive binding, we can reveal how loop size preferences vary across different targets. This unified platform accelerates radiopharmaceutical development by delivering both high-affinity ligands and the structural insights needed to optimize them for imaging applications.
Hypothesis/Objective:
Our objective was to develop and validate a unified platform combining variable-loop phage display with computational structure-function analysis to reveal how epitope characteristics dictate peptide architecture requirements, accelerating ligand discovery through structure-informed radiopharmaceutical design.
Methods:
We constructed a high-diversity (>10^8) disulfide-constrained peptide library with systematic loop-length variation (0-13 residues) that enhances proteolytic resistance and improves in vivo pharmacokinetics. To establish a baseline for distinguishing phage productionrelated biases from target-driven selection, we sequenced and compared loop-length distributions in both the DNA library and phage-displayed populations. The library was screened against monoclonal antibodies, purified protein targets, and whole cells using established human multiple myeloma and pancreatic cancer cell lines. We developed a next-generation sequencing (NGS) analysis pipeline incorporating machine learning-based denoising and motif discovery algorithms to efficiently distinguish true binders from false positives, enabling robust structure-function analysis across targets4.
Results:
Library characterization revealed minor loop length differences between DNA and phage-displayed populations, indicating phage production and amplification biases that may propogate over multiple rounds of screening. Proof-of-concept screens against monoclonal antibodies with known epitopes demonstrated two distinct patterns: i) antibodies that favor linear epitopes showed no loop length preference, and ii) antibodies targeting conformational epitopes exhibited clear loop length biases. Screens against various purified protein targets similarly revealed target-dependent loop length selection patterns, with different antigens favoring distinct structural architectures.
Conclusion:
We demonstrate that an integrated platform combining a variable-loop-length phage library with specialized NGS analysis reveals target-specific structural requirements for peptide binding. Our findings establish that protein targets impose distinct loop-length preferences: linear epitopes depend primarily on amino acid sequence and show no structural bias, while conformational epitopes impose clear architectural constraints. By quantifying these targetselective structural requirements and distinguishing them from structures favored during phage production, the platform provides mechanistic insights that inform rational radiopharmaceutical design. Ongoing validation and radiolabeling of candidates targeting multiple myeloma and pancreatic cancer will establish clinical utility5. This publicly accessible platform eliminates the need for multiple fixed-scaffold libraries and provides standardized computational tools, accelerating discovery of selective peptide-based imaging agents for precision cancer diagnostics and therapy.
Image/Figure:
Click to view full size
Click to view full size
Image/Figure Caption:
Overview of the phage display data processing and analysis pipeline for structural binding discovery. NGS datasets are decoded into amino acid sequences, error-corrected, and exported for analysis. Ligands are ranked using screen variables and a machine learning model, then clustered by sequence motif to identify distinct binding patterns. Applied to a structurally diverse variable-loop library, this pipeline enables readout of target-imposed loop and structural requirements while maximizing hit discovery across a broad chemical space.
Author
Stephen Lees
University of Virginia