Yang Li-Yen, Hicks Daniel J, Russo Paul S, McShan Andrew C
School of Chemistry and Biochemistry, Georgia Institute of Technology, Atlanta, Georgia, USA.
School of Materials Science and Engineering, Georgia Institute of Technology, Atlanta, Georgia, USA.
Protein Sci. 2025 Sep;34(9):e70279. doi: 10.1002/pro.70279.
Hydrophobins are a family of small fungal proteins that self-assemble at hydrophobic-hydrophilic interfaces. Hydrophobins not only play crucial roles in filamentous fungal growth and development but also have attracted substantial attention due to their unique material properties. Structural characterization of class I and class II hydrophobins to date has been limited to a handful of proteins. While machine-learning-based structure prediction methods have the potential to exponentially expand our ability to define global structure-function relationships of biomolecules, they have not yet been extensively applied to hydrophobins. Here, we apply a suite of bioinformatics tools including Rosetta, AlphaFold, FoldMason, and Foldseek toward analysis, modeling, classification, and global comparison of class I and class II hydrophobins. We first probe the structural and energetic features of experimental class I and class II structures available in the Protein Data Bank. Using previously solved X-ray and NMR structures, we benchmark the ability of AlphaFold to predict class I and class II hydrophobin folds. We explore the physicochemical properties of more than 7,000 class I and class II hydrophobins in the UniProt database. Then, using AlphaFold models, we classify the structural universe of all known class I and class II hydrophobins into six distinct clades. We also uncover putative non-canonical features of hydrophobins, including extended N-terminal tails, five disulfide bonds, polyhydrophobins, and non-hydrophobin proteins containing hydrophobin-like folds. Finally, we examine the ability of AlphaFold and Chai-1 to model hydrophobin membrane binding, conformational changes, and self-assembly of class I rodlets and class II meshes. Together, our results highlight that AlphaFold not only accurately models and enables the global comparison of features within the hydrophobin protein family but also uncovers new properties that can be further evaluated with experimentation.
疏水蛋白是一类小的真菌蛋白,它们在疏水-亲水界面处自组装。疏水蛋白不仅在丝状真菌的生长和发育中起着关键作用,而且因其独特的材料特性也引起了广泛关注。迄今为止,I类和II类疏水蛋白的结构表征仅限于少数几种蛋白质。虽然基于机器学习的结构预测方法有可能成倍扩展我们定义生物分子全局结构-功能关系的能力,但它们尚未广泛应用于疏水蛋白。在这里,我们应用了一套生物信息学工具,包括Rosetta、AlphaFold、FoldMason和Foldseek,来分析、建模、分类和全局比较I类和II类疏水蛋白。我们首先探究了蛋白质数据库中可用的实验性I类和II类结构的结构和能量特征。利用先前解析的X射线和核磁共振结构,我们对AlphaFold预测I类和II类疏水蛋白折叠的能力进行了基准测试。我们探索了UniProt数据库中7000多种I类和II类疏水蛋白的物理化学性质。然后,利用AlphaFold模型,我们将所有已知的I类和II类疏水蛋白的结构全域分类为六个不同的进化枝。我们还发现了疏水蛋白的推定非典型特征,包括延长的N末端尾巴、五个二硫键、多疏水蛋白以及含有疏水蛋白样折叠的非疏水蛋白。最后,我们研究了AlphaFold和Chai-1对I类棒状体和II类网格状结构的疏水蛋白膜结合、构象变化和自组装进行建模的能力。总之,我们的结果表明,AlphaFold不仅能准确地对疏水蛋白家族内的特征进行建模并实现全局比较,还能揭示新的特性,可通过实验进一步评估。