NMR Fingerprinting: Leveraging optimal control pulse design, tailored isotope labeling, and machine learning to study intractable proteins

NMR 指纹图谱：利用最佳控制脉冲设计、定制同位素标记和机器学习来研究棘手的蛋白质

• 批准号: 10387787
• 负责人: Haribabu Arthanari
• 金额: $ 7.5万
• 依托单位: DANA-FARBER CANCER INST
• 依托单位国家: 美国
• 财政年份: 2020
• 资助国家: 美国
• 起止时间: 2020-06-01 至 2025-03-31
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/10387787
• 关键词: Algorithms; Amino Acid Sequence; Amino Acids; Biochemical; Biophysics; Cell Nucleus; Computational Technique; Crowding; Data; Environment; Fingerprint; G-Protein-Coupled Receptors; Goals; Isotope Labeling; Isotopes; Machine Learning; Mathematics; Medical Research; Methods; Mission; NMR Spectroscopy; Nuclear Magnetic Resonance; Outcome; Pattern; Physiologic pulse; Positioning Attribute; Protein Dynamics; Proteins; Research; Resolution; Sampling; Science; Shapes; Signal Transduction; Specialist; Spectrum Analysis; Structure; System; Testing; Therapeutic Studies; artificial neural network; automated analysis; design; innovation; insight; next generation; protein function; quantum; radio frequency; therapeutic development

Project summary Nuclear magnetic resonance (NMR) spectroscopy is essential for the study structure, dynamics and function of proteins in near-native conditions. NMR studies have vital implications for therapeutic development. However, as the number of amino acids in the protein increases, NMR signals decay (relax) faster, yielding lower sensitivity and resolution, while the spectrum becomes more crowded. In these cases it is challenging to match observed signals to specific nuclei in the protein (called `resonance assignment') in order to meaningfully interpret NMR data. The overarching goal of our research is to push the boundaries of NMR enabling valuable insight about the dynamics and functions of currently intractable proteins. The objective of this project is to design an NMR platform consisting of coordinated, next-generation biochemical, biophysical, mathematical, and computational techniques. Our platform is built around an original approach to NMR spectroscopy in which new information about the local environment of each nucleus is encoded in the shape and pattern of its NMR signal. The rationale is that these patterns are a `fingerprint' – an intricate and unique signature that encodes key information about which atom is responsible for each resonance peak in the NMR spectrum. We will design and realize fingerprint patterns using two innovative approaches: 1) biochemically, by selectively introducing NMR-active isotopes into carefully chosen positions in the protein samples, and biophysically, and 2) by using specialized radiofrequency pulses to accurately control the quantum interactions that determine the NMR spectrum. The resulting fingerprints will be decoded using established algorithmic structures from machine learning, notably artificial neural networks. This will facilitate automated analyses that are accessible to non-NMR specialists. Our approach to spectroscopy holds promise in the study of therapeutically important proteins expressed in eukaryotic expression systems (e.g. G-protein coupled receptors and glycosylated proteins). Current NMR data from such proteins shows clear dynamics and interactions with other proteins, but cannot yet be properly interpreted because of the difficulty of relating each NMR peak to an amino acid in the protein sequence. Our platform will deliver two significant outcomes: 1) NMR resonance assignment for meaningful analyses of previously intractable systems. 2) Enable non-NMR specialists, to easily proceed from expressing their protein sample to using NMR to study dynamics and interactions via assigned spectra. This will have a positive impact on protein science and medical research. To support our mission we have assembled a team of leading experts to test our platform with their own protein systems.

项目概要 核磁共振 (NMR) 光谱对于研究结构、动力学和功能至关重要 接近天然条件下的蛋白质。核磁共振研究对治疗开发具有重要意义。然而， 随着蛋白质中氨基酸数量的增加，NMR 信号衰减（松弛）得更快，产量更低 灵敏度和分辨率，而光谱变得更加拥挤。在这些情况下，匹配是具有挑战性的 观察到蛋白质中特定核的信号（称为“共振分配”），以便有意义地 解释 NMR 数据。我们研究的首要目标是突破 NMR 的界限，实现有价值的 关于当前棘手蛋白质的动力学和功能的见解。该项目的目标是 设计一个由协调的下一代生物化学、生物物理、数学、 和计算技术。我们的平台是围绕核磁共振波谱的原始方法构建的，其中 有关每个原子核局部环境的新信息被编码在其 NMR 的形状和模式中 信号。基本原理是这些模式是“指纹”——一种复杂而独特的签名，用于编码 关于核磁共振谱中每个共振峰由哪个原子负责的关键信息。我们将 使用两种创新方法设计和实现指纹图案：1）生物化学，选择性地 将 NMR 活性同位素引入蛋白质样品中精心选择的位置，并通过生物物理方法， 2）通过使用专门的射频脉冲来精确控制决定量子相互作用的量子相互作用 核磁共振谱。生成的指纹将使用已建立的算法结构进行解码 机器学习，特别是人工神经网络。这将促进可访问的自动化分析 非 NMR 专家。我们的光谱学方法有望在治疗上重要的研究中发挥作用 在真核表达系统中表达的蛋白质（例如 G 蛋白偶联受体和糖基化 蛋白质）。目前此类蛋白质的 NMR 数据显示出清晰的动力学以及与其他蛋白质的相互作用，但是 由于很难将每个 NMR 峰与氨基酸中的氨基酸相关联，因此尚无法正确解释 蛋白质序列。我们的平台将带来两个重要成果：1) NMR 共振分配 对以前棘手的系统进行有意义的分析。 2) 使非 NMR 专家能够轻松地从 使用 NMR 表达蛋白质样品，通过指定的光谱研究动力学和相互作用。这 将对蛋白质科学和医学研究产生积极影响。为了支持我们的使命，我们有 组建了一支领先的专家团队，用他们自己的蛋白质系统测试我们的平台。