Abstract 4225: AI-driven de novo design of humanized nanobodies targeting KLK2 for prostate cancer immunotherapy

Abstract Conventional monoclonal antibodies (mAbs) are widely used in cancer therapy but are limited by large size (∼150 kDa), structural complexity, and high production costs, motivating the development of alternative antibody formats. Single-chain variable fragments (scFvs), derived from mAbs, retain antigen-binding specificity and affinity while offering a smaller molecular size (∼25 kDa), enabling applications such as chimeric antigen receptor (CAR)-T/NK cells and/or bispecific T/NK cell engagers (BiTEs or NKCEs) for immunotherapy. However, scFvs face challenges including structural instability, linker-dependent functionality, and aggregation-prone hydrophobic residues, which limit their therapeutic efficacy. In contrast, single-domain antibodies (sdAbs), derived from the variable antigen-binding domain of heavy-chain-only antibodies (VHHs) and commonly known as nanobodies (∼15 kDa), overcome these limitations. Importantly, nanobodies can access cryptic or conformational epitopes—such as deep cavities that are inaccessible to conventional mAbs or scFvs—providing a stable, highly versatile, and clinically attractive biotherapeutic modality for cancer immunotherapy. Traditional methods to develop nanobodies rely on targeting protein-immunized camelids or high-throughput experimental nanobody display systems, including bacteria, phage, ribosome, or yeast display. These approaches are time-consuming, labor-intensive, camelid-derived (with potential immunogenicity), and restricted in library diversity. Moreover, affinity maturation and specificity optimization require multiple iterative rounds of experimental selection—often taking months to years—and may still fail to produce high-affinity nanobodies against difficult targets. To address these challenges, we developed and integrated multiple artificial intelligence (AI) platforms for target-guided humanized nanobody backbone design, soluble protein sequence generation, and enhanced target-nanobody complex prediction and screening. These platforms enable de novo, conformational epitope-based nanobody design, generating diverse, high-affinity, and highly specific candidates entirely in silico, with experimental validation achievable within weeks to months depending on the target. As a proof-of-principle, we have applied our AI platforms to design humanized nanobodies targeting kallikrein-related peptidase 2 (KLK2), a clinically validated, prostate cancer-specific cell surface protein, and are experimentally engineering the top-designed candidates into T/NK cell engagers and CAR-T/NK cells to evaluate their ability to eliminate KLK2-expressing prostate cancer cells. Citation Format: Fengze Jin, Puja Singh, Hanyong Chen, Christopher Warlick, Yibin Deng, . AI-driven de novo design of humanized nanobodies targeting KLK2 for prostate cancer immunotherapy 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 4225.