What question did this study set out to answer?

The research aims to develop an AI-integrated platform for efficient discovery and validation of cyclic peptide ligands targeting B7-H3 and DLL3.

April 5, 2026

Abstract 6409: AI-accelerated discovery of B7-H3 and DLL3-targeted cyclic peptide radioligands: From library design to preclinical validation

Key Points

The research aims to develop an AI-integrated platform for efficient discovery and validation of cyclic peptide ligands targeting B7-H3 and DLL3.
Engineered a phage display library with extensive diversity for peptide discovery.
Utilized deep sequencing and AlphaFold3 for structural modeling of peptide-target interactions.
Conducted SPR and spectral shift assays for binding affinity assessments and selectivity testing.
15 out of 40 phage clones showed significant binding to target proteins.
AlphaFold3 predictions identified a consistent β-turn motif in 7 of 15 effective hits.
Synthesized peptides demonstrated a binding affinity of KD 8.2×10-7 M with over three-fold selectivity against similar proteins.

Abstract

Abstract Background: Radiopharmaceutical drug conjugates (RDCs) represent a transformative paradigm in precision oncology, yet target-specific ligand discovery remains a critical bottleneck. We developed an integrated AI-augmented platform to rapidly identify cyclic peptide binders for two emerging RDC targets: B7-H3 (immune checkpoint) and DLL3 (Notch ligand). Methods: A structurally diverse phage display library (1.5×1011capacity; 8-17 aa macrocycles) was engineered with NGS-validated complexity. Recombinant 4Ig-B7-H3 (2Ig-B7-H3 as off-target counterscreens), DLL3 (and counterscreens DLL1/DLL4) underwent orthogonal biophysical characterization by SPR and Spectral shift assay (SPS). Hit triangulation employed: (1) Deep sequencing-driven consensus motif analysis, (2) AlphaFold3 multimer modeling of peptide-target complexes, and (3) Parallel SPR/spectral shift assays (10-7 M affinity threshold). Top candidates were Cy5-labeled for real-time binding and internalization kinetics in engineered tumor lines. Results: Discovery: 15/40 phage clones demonstrated target binding - AI optimization: AlphaFold3 predictions revealed a conserved β-turn motif in 7/15 top hits that anchors to a cryptic pocket in targets - Validation: Hit peptide were synthesized and showed: (i) KD 8.2×10-7 M (SPR), (ii) 3-fold selectivity over homologous target proteins (iii) showed binding on cells Conclusions: This platform addresses critical challenges in RDC development by providing a streamlined solution for hit discovery, optimization, and validation. The integration of AI-driven structural prediction with high-throughput experimental validation compresses traditional hit discovery timelines This platform will significantly accelerate the RDC development pipeline—especially for hit binder identification for neo tumor antigens or other validated surface proteins like GPCRs, transporters, to accelerate RDC drug discovery using cyclic peptide as the modality. Citation Format: Tj (Tiejun) Bing. AI-accelerated discovery of B7-H3 and DLL3-targeted cyclic peptide radioligands: From library design to preclinical validation abstract. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 6409.

Bookmark

Abstract 6409: AI-accelerated discovery of B7-H3 and DLL3-targeted cyclic peptide radioligands: From library design to preclinical validation

Key Points

Abstract

Cite This Study

Also Consider

Also Consider