LBPO.CL01 · 临床研究 · Late-Breaking
整合计算设计与cDNA展示筛选实现环肽放射性药物配体的发现
Integrated computational design and cDNA display selection enable cyclic peptide radiopharmaceutical ligand discovery
作者与单位 Authors & Affiliations
摘要 Abstract
中文摘要
尽管尺寸小，环肽在结合亲和力和生物特异性之间提供了理想的平衡，使其成为放射性药物和基于肽的诊疗一体化(theranostics)有吸引力的支架。然而，由于设计和筛选策略有限，鉴定适合正电子发射断层扫描(PET)成像或放射治疗有效载荷递送的新型环肽配体仍是一项挑战。为满足这一需求，我们开发了一个整合流程，将大环肽的计算设计与通过cDNA展示的高通量文库筛选相结合。我们的计算工作流首先查询结构数据库，寻找将选定表位基序作为亚结构包含的蛋白质。随后我们从这些蛋白质中提取相互作用的三级基序(TERM)作为设计种子，反映天然蛋白质的形状互补性和界面特征。使用这些TERM，我们用Protpardelle生成多样化的肽骨架集合，Protpardelle是一个通过独特的偏置交换元动力学能力在硫醚闭合环肽的分子动力学模拟上训练的生成模型。候选支架接下来使用ProteinMPNN和Rosetta进行序列设计，随后进行全位点饱和突变以系统探索有利于亲和力优化的有益突变。随后的cDNA展示筛选和二代测序(NGS)实现了对富集的高亲和力结合物的定量鉴定和排序。我们已成功应用该平台设计靶向细胞表面蛋白的环肽，包括免疫检查点调节因子和泛癌标志物，并将在此呈现跨B7家族靶点的初步数据。这些靶点在实体瘤以及乳腺癌、卵巢癌和子宫内膜癌的肿瘤相关巨噬细胞中过表达，已成为高价值的治疗抗原。虽然针对B7-H3/B7-H4的抗体药物偶联物(ADC)正在临床开发中，但其较大的分子尺寸和缓慢的清除会限制肿瘤穿透并导致剂量限制性毒性。相比之下，紧凑的环肽表现出快速的肿瘤摄取和清除，这些特性非常适合PET成像和靶向放射性核素治疗。通过将基序引导的计算设计与基于NGS读出的筛选相结合，我们的框架加速了可供合成和放射性标记的高亲和力环肽的发现，从而推动基于肽的诊疗一体化走向临床转化。
查看英文原文 English abstract
Despite their small size, cyclic peptides offer a desirable balance of binding affinity and biological specificity, making them attractive scaffolds for radiopharmaceuticals and peptide-based theranostics. However, identifying novel cyclic peptide ligands suitable for positron emission tomography (PET) imaging or radiotherapeutic payload delivery remains a challenge due to limited design and screening strategies. To address this need, we developed an integrated pipeline that combines computational design of macrocyclic peptides with high-throughput library screening via cDNA display. Our computational workflow begins by querying structural databases for proteins containing a selected epitope motif as a substructure. We then extract interacting tertiary motifs (TERMs) from these proteins as design seeds that reflect natural protein shape complementarity and interface features. Using these TERMs, we generate diverse peptide backbone ensembles with Protpardelle, a generative model trained on molecular dynamics simulations of thioether-closed cyclic peptides through a unique bias exchange metadynamics capability. Candidate scaffolds are next sequence-designed using ProteinMPNN and Rosetta, followed by full site-saturation mutagenesis to systematically explore advantageous mutations for affinity optimization. Subsequent cDNA display selections and next-generation sequencing (NGS) enable quantitative identification and ranking of enriched, high-affinity binders. We have successfully applied this platform to design cyclic peptides targeting cell-surface proteins, including immune checkpoint regulators and pan-cancer markers and here will present preliminary data across the B7 family of targets. These targets are overexpressed across solid tumors and tumor-associated macrophages in breast, ovarian, and endometrial cancers and have emerged as high-value therapeutic antigens. While antibody-drug conjugates (ADCs) against B7-H3/B7-H4 are in clinical development, their large molecular size and slow clearance can limit tumor penetration and contribute to dose-limiting toxicity. In contrast, compact cyclic peptides exhibit rapid tumor uptake and clearance, properties well-suited for PET imaging and targeted radionuclide therapy. By combining motif-guided computational design with selection-based screening readout by NGS, our framework accelerates the discovery of high-affinity cyclic peptides ready for synthesis and radiolabeling, thereby advancing peptide-based theranostics toward clinical translation.
利益披露 Disclosure
J. L. Fieger, None..
B. V. Bell, None.