The regulation of various cellular processes heavily relies on the protein
complexes within a living cell, necessitating a comprehensive understanding of
their three-dimensional structures to elucidate the underlying mechanisms.
While neural docking techniques have exhibited promising outcomes in binary
protein docking, the application of advanced neural architectures to multimeric
protein docking remains uncertain. This study introduces SyNDock, an automated
framework that swiftly assembles precise multimeric complexes within seconds,
showcasing performance that can potentially surpass or be on par with recent
advanced approaches. SyNDock possesses several appealing advantages not present
in previous approaches. Firstly, SyNDock formulates multimeric protein docking
as a problem of learning global transformations to holistically depict the
placement of chain units of a complex, enabling a learning-centric solution.
Secondly, SyNDock proposes a trainable two-step SE(3) algorithm, involving
initial pairwise transformation and confidence estimation, followed by global
transformation synchronization. This enables effective learning for assembling
the complex in a globally consistent manner. Lastly, extensive experiments
conducted on our proposed benchmark dataset demonstrate that SyNDock
outperforms existing docking software in crucial performance metrics, including
accuracy and runtime. For instance, it achieves a 4.5% improvement in
performance and a remarkable millionfold acceleration in speed.

本研究介绍了 SyNDock，一种自动化框架，可以在几秒钟内快速组装精确的多聚体复合物，显示出在关键性能指标方面优于现有的对接软件的性能，并将多聚体蛋白对接表述为学习全局转换的问题，其可应用于将链单元置于全局一致性的解决方案中，同时提出了一种可训练的两步 SE（3）算法。