Development of a pipeline for parallel elucidation of protein structures

生物物理技术的进步加快了我们探索即使是最严重的 复杂的细胞系统，这样的研究使研究人员能够设计对已知蛋白质的修饰 构建和设计全新的蛋白质。这种“蛋白质设计”技术已经产生了一种 操纵蛋白质结构作为改进或引入新的医学诊断和 治疗学。这些研究的基础依赖于候选蛋白质的计算模型，尽管 蛋白质结构预测、蛋白质从头设计和单一突变效应预测的准确性仍然存在。 低于许多用例的门槛，例如结构指导的药物设计和合理的酶工程。 因此，蛋白质工程的成功依赖于高分辨率的结构确定，这涉及到 为了获得稳定的蛋白质或活性酶变异体，进行了艰苦的筛选和优化。然而，我们的 使用普通结构测定策略观察蛋白质结构的能力(X射线结晶学， 核磁共振和低温电子显微镜(CRYO-EM))远远落后于我们设计和生产新序列的能力， 造成了一种知识鸿沟，阻止了生物化学家获得自然界中看到的一系列蛋白质功能。 虽然目前的技术能够快速合成数百种具有不同序列的蛋白质，但还没有 现有的技术可以快速表征这些产生的蛋白质的结构。能够获得高度的- 数百个并行序列的解析结构信息将为蛋白质提供无价的洞察力 工程学方法。重要的是，快速结构确定将使结构表征成为可能 人类基因组中的遗传变异使结构性和机械性疾病成为可能 罕见和新发疾病相关变异的解释。Cryo-EM实现了众多高分辨率 从少量样品中确定结构而不要求均一性，这是这方面的一个方面 我们计划用来平行阐明蛋白质结构的方法。我们将确定这一点的可行性 用于快速研究工程蛋白质文库结构的技术，其中，分子量范围 接近或低于低温EM的检测下限。我们还将探索我们识别 在有限的结构位置，如活性部位测试突变的位置和结构影响。我们会 在两个目标中探索我们的并行结构确定方法的可行性：目标1将确定 目前用于区分结构相似的蛋白质复合体的单颗粒分析方法。目标2 将实施机器学习算法，以使用组合的 人工合成的真实数据。这些探索性研究将为快速确定多个化合物的结构奠定基础 来自单一低温EM实验的蛋白质复合体，提供快速获得高分辨率的能力 许多工程蛋白质的结构，从而实现前所未有的设计和测试反馈周期 帮助治疗人类疾病。

Advances in biophysical technologies have accelerated our ability to probe the mechanisms of even the most complex cellular systems, and such studies have enabled researchers to design modifications to known protein structures and design completely new proteins. This “protein design” technology has given rise to an ability to manipulate protein structures as a means of improving on or introducing new medical diagnostics and therapeutics. The bases of these studies rely on computational modeling of protein candidates, although the accuracy of protein structure prediction, protein de novo design, and single-mutation effects prediction remain below the threshold for many use cases, such as structure-guided drug design and rational enzyme engineering. Thus, success of a protein engineering effort relies on high-resolution structure determination, which involves laborious screening and optimization in order to obtain stable proteins or active enzyme variants. However, our ability to observe protein structure using common structure determination strategies (X-ray crystallography, NMR, and cryo-electron microscopy (cryo-EM)) lags far behind our ability to design and produce new sequences, creating a knowledge gap that prevents biochemists from accessing the range of protein functions seen in nature. While current technologies enable rapid synthesis of hundreds of proteins with varied sequences, there do not exist technologies for rapid structural characterization of these generated proteins. The ability to obtain high- resolution structural information for hundreds of sequences in parallel would provide invaluable insights in protein engineering methods. Importantly, rapid structure determination would enable structural characterization of genetic variation in the human genome underlying disease by enabling the structural and mechanistic interpretation of rare and de novo disease-related variants. Cryo-EM enables numerous high-resolution structures to be determined from a small amount of sample without requiring homogeneity, an aspect of this method that we plan to exploit for parallel elucidation of protein structures. We will establish the feasibility of this technique for rapidly investigate the structures of engineered protein libraries, where the molecular weight range is near or below the lower detection limit of cryo-EM. We will also probe the limits of our ability to identify the location and structural impact of tested mutations at limited structural locations, such as active sites. We will explore the feasibility of our parallel structure determination approach in two aims: Aim 1 will identify the limit of current single-particle analysis methods to discriminate between structurally similar protein complexes. Aim 2 will implement machine learning algorithms to push the current limits of classification using a combination of synthetic and real data. These exploratory studies will pave the way to rapid structure determination of multiple protein complexes from a single cryo-EM experiment, providing the ability to rapidly obtain high-resolution structures for many engineered proteins, thereby enabling unprecedented design and testing feedback cycles to help treat human disease.