收录解读
膜蛋白的结构解析和疫苗开发因其广泛的疏水表面而面临巨大挑战。本文提出了一种基于深度学习的通用设计方法，用于增溶天然膜蛋白，同时保留其序列、折叠、活性位点和配体结合特性。
核心方法是利用深度学习（RFdiffusion）设计基因编码的de novo蛋白——WRAP（水溶性RFdiffused两亲性蛋白），它们包裹膜蛋白的脂质相互作用疏水表面，使其无需去污剂即可保持热稳定和水溶性。
实验验证包括：对单体/寡聚β-桶外膜蛋白和螺旋多次跨膜蛋白成功设计WRAPs；解析了2.95埃分辨率的冷冻电镜结构；展示了WRAPs可用于膜蛋白的结构测定。此外，应用于梅毒疫苗开发，生成了梅毒螺旋体抗原的可溶版本。
该方法为膜蛋白研究提供了可复用的通用平台，有望推动结构生物学和疫苗设计领域的发展。局限性在于目前仅验证了几类膜蛋白，尚未覆盖所有类型，但已展示出广泛的适用性和可扩展性。
原始摘要
Developing therapies and vaccines against integral membrane proteins is hindered by their extensive hydrophobic surfaces, which complicate production and structural analysis. Here, we describe a general deep learning–based design approach for solubilizing native membrane proteins while preserving their sequence, fold, active-site, and ligand-binding properties. Genetically encoded de novo protein WRAPs [water-soluble RFdiffused amphipathic proteins] surround the lipid-interacting hydrophobic surfaces, rendering them thermostable and water-soluble without the need for detergents. We design WRAPs for both monomeric and oligomeric beta-barrel outer membrane proteins and helical multipass transmembrane proteins. A 2.95-angstrom-resolution cryo–electron microscopy structure of WRAPed mycobacterial porin demonstrates that WRAPs can be used for the structural determination of membrane proteins in solution. As a step toward syphilis vaccine development, we generated soluble versions of Treponema pallidum antigens.