发表机构
Tsinghua University; Zhejiang University; Shanghai University; Xi’an Jiaotong University（清华大学; 浙江大学; 上海大学; 西安交通大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
Hyper-Fold通过超图建模探索蛋白质序列-几何学习的表达极限，其相关变体在多项蛋白质结构相关任务中取得最优结果，且Hyper-Fold-Pocket性能优于对比模型，参数与延迟更优。
AI 中文摘要
蛋白质结构建模依赖单一计算基元：残基的序列内容（即残基是什么）与其三维几何位置（即残基在哪里）之间的交互。这类层的表达极限是什么？研究表明，内容-几何外积上的完全双线性算子（所有二阶交互的充分统计量）是表达上限，而主流几何图神经网络（GNN）的加性消息传递对内容-几何绑定具有不可知性。随后引入Hyper-Fold，这是一种秩为K的可分离卷积骨干网络，以消息传递的成本逼近该上限：每个半径邻域被组织为序列超边和接触超边，由边缘条件矩阵值算子调制，该算子被分解为K个带几何生成系数的学习基算子。在酶功能预测、折叠分类和配体结合位点检测任务中，Hyper-Fold及其分层变体Hyper-Fold-Deep在蛋白质特异性结构编码器中取得最优结果；Hyper-Fold-Pocket作为锚定集预测头，在不使用序列语言模型特征的情况下，在UniSite-DS和两个零样本基准上超越UniSite-3D，参数减少68倍，延迟降低4.8倍——这表明足够具表达力的3D骨干网络可恢复融合架构此前从进化规模预训练中获取的信息。
英文摘要
Protein structure modeling rests on a single computational primitive: the interaction between what a residue is (sequence content) and where it sits (three-dimensional geometry). What is the expressive limit of this layer class? We show that the complete bilinear operator over content-geometry outer products--the sufficient statistic of all second-order interactions--is the expressive ceiling, while the additive message passing of mainstream geometric GNNs is provably blind to content-geometry binding. We then introduce Hyper-Fold, a rank-K separable convolutional backbone approaching this ceiling at message-passing cost: each radius neighborhood is organized into a sequence hyperedge and a contact hyperedge, modulated by an edge-conditioned matrix-valued operator factorized into K learned basis operators with geometry-generated coefficients. Across enzyme function prediction, fold classification, and ligand binding site detection, Hyper-Fold and its hierarchical variant Hyper-Fold-Deep achieve the best results among protein-specific structure encoders; Hyper-Fold-Pocket, an anchored set-prediction head, surpasses UniSite-3D on UniSite-DS and two zero-shot benchmarks with no sequence language model features, 68x fewer parameters, and 4.8x lower latency--suggesting that a sufficiently expressive 3D backbone recovers information that fusion architectures previously borrowed from evolution-scale pretraining.
