Computational design of protein-binding proteins is a fundamental capability with broad utility in biomedical research and biotechnology. Recent methods have made strides against some target proteins, but on-demand creation of high-affinity binders without multiple rounds of experimental testing remains an unsolved challenge. This technical report introduces AlphaProteo, a family of machine learning models for protein design, and details its performance on the de novo binder design problem. With AlphaProteo, we achieve 3- to 300-fold better binding affinities and higher experimental success rates than the best existing methods on seven target proteins. Our results suggest that AlphaProteo can generate binders "ready-to-use" for many research applications using only one round of medium-throughput screening and no further optimization.
核心贡献 · Key contributions
提出 AlphaProteo，一个用于从头设计蛋白质结合体的机器学习模型家族。 Introduces AlphaProteo, a family of ML models for de novo protein binder design.
在七个靶标上实现比现有方法高 3 到 300 倍的结合亲和力。 Achieves 3- to 300-fold better binding affinities than existing methods on seven targets.
仅通过一轮中等通量筛选即展示出高实验成功率（9%-88%）。 Demonstrates high experimental success rates (9%-88%) with only one round of medium-throughput screening.
对大多数靶标无需亲和力优化即获得亚纳摩尔至低纳摩尔级别的结合体。 Obtains sub-nanomolar to low-nanomolar binders without affinity optimization for most targets.
验证生物学活性：抑制 VEGF 信号传导和中和 SARS-CoV-2。 Validates biological activity: inhibition of VEGF signaling and SARS-CoV-2 neutralization.
通过冷冻电镜和 X 射线晶体学高精度确认设计结构。 Confirms designed structures via cryo-EM and X-ray crystallography with high accuracy.
局限 · Limitations
仅在八个靶标蛋白上验证；对更广泛靶标的泛化性需进一步测试。 Validated on only eight target proteins; generalization to broader targets needs further testing.
由于平坦极性结合位点，未能获得一个挑战性靶标（TNFα）的结合体。 Failed to obtain binders for one challenging target (TNFα) due to flat polar binding site.
需要靶标晶体结构作为输入；对缺乏实验结构的靶标有限制。 Requires target crystal structure as input; limited for targets without experimental structures.
由于生物安全和商业考虑，未公开机器学习方法。 Machine learning methods not disclosed due to biosecurity and commercial considerations.
结合体仅设计用于研究用途，不用于临床应用。 Binders designed for research use only, not for clinical applications.
论文章节 · Sections（共 9）
摘要Abstract
1 引言1 Introduction
2 结果2 Results
2.1 中通量筛选获得亚纳摩尔亲和力结合剂2.1 Sub-nanomolar-affinity binders from medium-throughput screening
2.2 结合剂的功能与结构验证2.2 Functional and structural validation of binders