Towards Accurate Protein Structure Predictions with SAXS Technology (TAPESTRY)

项目摘要 在医学和基础科学中，对蛋白质的精确原子结构存在未满足的需求。这种需要 超越了传统高分辨率实验方法的能力。随着机器学习的进步， 结构预测算法准备在不久的将来为这些领域提供原子模型。然而 预测算法中的差距限制了准确性和可靠性，特别是对于大的多结构域蛋白质、蛋白质 复合物和柔性蛋白质。我们的建议，用SAXS精确预测蛋白质结构 技术（TAPESTALY），将创造技术，以提高可靠性和提高蛋白质的准确性 通过实验验证进行结构预测，特别是对于困难的蛋白质。 TAPEscore是创新的，结合了我们在高通量同步加速器SAXS（小角度）方面的优势 X射线散射）数据收集和分析，以及蛋白质结构的关键评估 预测（CASP），根据“黄金标准”评估结构预测，尚未发布晶体 每两年建造一次。通过CASP，我们利用集体蛋白质折叠知识， 结构预测科学家的全球社区。 我们的方法是战略性的。我们为CASP提供SAXS数据，使预测科学家能够访问 实验数据我们开发分析和实验工具，专为预测科学家设计， 克服目前限制使用SAXS数据的差距。我们根据我们的TAPE数据库测试这些工具 标准蛋白质，与相应的晶体结构，SAXS数据和预测模型。最后我们 通过CASP评估我们技术的稳健性，并获得对我们工具的公正评估 和磁场的状态作为第一步，我们的目标是折叠良好的蛋白质（Aim 1）和具有无序尾部的蛋白质 (Aim（2）在本提案中。 我们的技术提案的可行性得到了我们现有数据和概念验证的支持，我们的 光束线能力和在SAXS分析方面的成熟经验。我们表明，实验SAXS数据， 它包含距离信息，可以作为蛋白质结构预测算法的约束，匹配 良好折叠蛋白质的晶体结构和基于拓扑准确性的得分预测。我们展示案例 在CASP13（2018）中，SAXS数据改善了预测模型的倍数。SAXS数据收集迅速（10 秒），不需要标记或结晶，并且科学界可以免费获得。我们 在开发信息丰富和有效的SAXS分析工具方面拥有丰富的经验。 我们的长期目标是使生物医学研究人员能够输入氨基酸序列，并迅速获得一个氨基酸序列。 反映溶液中蛋白质构象的实验验证和精确的原子模型。如果 如果TAPE成功，这种原子模型的可用性将有强大而广泛的潜力， 推进生物医学研究并影响蛋白质参与的所有生物学领域。

Project Summary There is an unmet need in medicine and basic sciences for accurate atomic structures of proteins. This need surpasses the capabilities of traditional high-resolution experimental methods. With machine learning advances, structure prediction algorithms are poised to provide atomic models for these areas in the near future. Yet, the gaps in prediction algorithms limit accuracy and reliability, particularly for large multi-domain proteins, protein complexes, and flexible proteins. Our proposal, Towards Accurate protein structure Predictions with SAXS TechnologY (TAPESTRY), will create technology to increase reliability and improve accuracy of protein structure predictions through experimental validation, particularly for difficult proteins. TAPESTRY is innovative by combining our strengths in high-throughput synchrotron SAXS (Small Angle X-ray Scattering) data collection and analysis with the Critical Assessment of protein Structure Prediction (CASP), which assesses structure predictions against “gold standard”, not-yet-released crystal structures every two years. Through CASP, we take advantage of the collective protein folding knowledge of the global community of structure prediction scientists. Our approach is strategic. We provide SAXS data for CASP, giving prediction scientists access to experimental data. We develop analytical and experimental tools, designed for prediction scientists to overcome current gaps that limit the use of SAXS data. We test these tools against our TAPESTRY databases of standard proteins, with corresponding crystal structures, SAXS data, and predicted models. Finally, we evaluate the robustness of our technology through CASP and obtain an unbiased assessment of our tools and the state of the field. As a first step, we target well-folded proteins (Aim 1) and proteins with disordered tails (Aim 2) in this proposal. The feasibility of our technology proposal is supported by our current data and proofs-in-concepts, our beamline capabilities, and proven experience in SAXS analysis. We show that experimental SAXS data, which contains distance information that can act as restraints in protein structure prediction algorithms, match crystal structures of well-folded proteins and score predictions based on topological accuracy. We show cases in CASP13 (2018) when SAXS data improved the fold of predicted models. SAXS data collection is rapid (10 seconds), does not require labeling or crystallization, and is available at no cost to the scientific community. We have proven experience in developing informative and effective SAXS analytical tools. Our long-term goal is to enable biomedical researchers to input an amino acid sequence and rapidly obtain an experimentally validated and accurate atomic model(s) that reflects the protein conformation(s) in solution. If TAPESTRY is successful, the increased availability of such atomic models will have strong and broad potential to advance biomedical research and impact all areas of biology in which proteins are involved.