High-Confidence Protein–Ligand Complex Modeling by NMR-Guided Docking Enables Early Hit Optimization

Structure-based drug
design is an integral part of modern day drug
discovery and requires detailed structural characterization of protein–ligand
interactions, which is most commonly performed by X-ray crystallography.
However, the success rate of generating these costructures is often
variable, in particular when working with dynamic proteins or weakly
binding ligands. As a result, structural information is not routinely
obtained in these scenarios, and ligand optimization is challenging
or not pursued at all, representing a substantial limitation in chemical
scaffolds and diversity. To overcome this impediment, we have developed
a robust NMR restraint guided docking protocol to generate high-quality
models of protein–ligand complexes. By combining the use of
highly methyl-labeled protein with experimentally determined intermolecular
distances, a comprehensive set of protein–ligand distances
is generated which then drives the docking process and enables the
determination of the correct ligand conformation in the bound state.
For the first time, the utility and performance of such a method is
fully demonstrated by employing the generated models for the successful,
prospective optimization of crystallographically intractable fragment
hits into more potent binders.

基于结构的药物设计是现代药物发现不可或缺的核心组成部分，其实施依赖于对蛋白质-配体相互作用的详细结构表征，而该类表征最常用的技术手段为X射线晶体学（X-ray crystallography）。然而，获取此类蛋白质-配体共晶结构的成功率往往波动不定，在针对动态蛋白质或弱结合配体开展研究时尤为明显。因此，此类场景下通常无法常规获取目标结构信息，配体优化工作要么困难重重，要么完全无法推进，这成为化学骨架开发与多样性拓展的重大瓶颈。为突破这一局限，我们开发了一套稳健的核磁共振（Nuclear Magnetic Resonance）约束引导对接方案，用于构建高质量的蛋白质-配体复合物模型。该方案将高甲基化标记蛋白质的使用与实验测得的分子间距离相结合，生成一套完整的蛋白质-配体距离数据集，以此驱动对接进程，进而确定配体在结合状态下的正确构象。本研究首次将该方法生成的复合物模型应用于晶体学难以解析的片段命中化合物的前瞻性优化，成功将其升级为活性更强的结合配体，全面验证了该方法的实用性与性能。