Learning the Shared Latent Structure of Cryo-EM Inverse Problems
Understanding biomolecular function requires characterizing three-dimensional structure and conformational variability, for which cryogenic electron microscopy (cryo-EM) is a powerful tool. Cryo-EM produces noisy two-dimensional particle images in which molecular structure, particle pose, and imaging conditions are jointly encoded.
理解生物分子功能需要表征其三维结构和构象变异性，而冷冻电子显微镜（cryo-EM）正是实现这一目标的有力工具。冷冻电镜产生带有噪声的二维颗粒图像，其中分子结构、颗粒姿态和成像条件被共同编码。
Recovering biological information from these observations spans inverse problems including pose estimation, heterogeneity analysis, and reconstruction. These tasks are usually addressed independently despite shared protein structure and image formation, motivating a common representation across inverse problems.
从这些观测结果中恢复生物信息涉及姿态估计、异质性分析和重构等反问题。尽管这些任务共享蛋白质结构和图像形成过程，但通常被独立处理，这促使人们寻求一种跨反问题的通用表示方法。
Since biologically meaningful differences may be spatially localized, we study self-distillation with complementary global and masked patch objectives. Because experimental dataset discrimination can reflect acquisition-specific cues, we develop controlled synthetic and experimental evaluations and find that structural discrimination benefits from dense spatial features.
由于具有生物学意义的差异可能在空间上是局部的，我们研究了结合全局和掩码补丁目标的自蒸馏方法。由于实验数据集的区分可能反映了采集过程中的特定线索，我们开发了受控的合成和实验评估，并发现结构区分受益于密集的空间特征。
We introduce the EMber family of vision transformers, pre-trained on >50 million experimental particle images from 214 EMPIAR datasets together with simulated projections from 268,000 molecular assemblies. EMber improves upon existing methods at conformational and compositional discrimination by 14.85 and 14.26 percentage points, estimates conformational populations to within 1.59pp error with a linear probe, and produces particle assignments yielding reference-quality reconstructions in fewer epochs.
我们引入了 EMber 系列视觉 Transformer，它在来自 214 个 EMPIAR 数据集的超过 5000 万张实验颗粒图像以及 268,000 个分子组装体的模拟投影上进行了预训练。EMber 在构象和组成区分方面比现有方法分别提高了 14.85 和 14.26 个百分点，通过线性探针将构象群体估计误差控制在 1.59 个百分点以内，并能以更少的迭代次数产生实现参考质量重构的颗粒分配。
We demonstrate how to learn representations that capture biological information for use across multiple cryo-EM inverse problems.
我们展示了如何学习能够捕获生物信息并用于多种冷冻电镜反问题的表示方法。