Combined Automatic Assignment of NMR Spectra and Calculation of Macromolecular Structure by Self-Correcting Distance Geometry

该项目将提高一种新的计算工具的稳定性、可靠性和效率，即用于核磁共振波谱和三维结构计算相结合的自校正几何（SECODG）方法。在DIAMOD/NOAH软件包中实现的SECODG方法大大减少了从NMR数据生成蛋白质结构的时间。与以前的距离几何方法不同，这些方法是为一致的数据集设计的，该方法可以从包含误差的约束集中生成精确的结构。该项目将进一步改进该方法，使其适用于更大的蛋白质。几次测试表明，该方法可以处理真实数据，并且比人工交互光谱解释方法快得多。第一个测试是比较NOESY光谱自动结构计算的结果，这些结果已经用早期的人工方法分析了6种蛋白质，大小从40到135个氨基酸不等。自动化方法分配了70-80%的NOESY交叉峰，三维结构质量相似。在第二次测试中，NOAH/DIAMOD使用先前未解释的noasy和TOCSY光谱来自动计算蛋白异构体的三维结构束。需要稍加修改和一些人工辅助，以便程序能够处理大量缺失的化学变化。基于这一实践经验，该方法将通过优化容错目标函数，结合约束的灵敏度测试（包括峰分配方法中的线形信息）以及将程序套件与其他组的自动顺序分配程序相结合来改进该方法，以开发用于直接解释NMR光谱的全自动计算包。这个程序包将是一个强大的计算工具，以加快从核磁共振数据确定大分子结构。实验确定的三维结构是设计具有改进或新功能的新药和蛋白质的基础。结合能量最小化和蒙特卡罗模拟，它将有助于设计具有所需结构和新功能特性的蛋白质。

This project will enhance the stability, reliability and efficiency of a new computational tool, the self- correcting geometry (SECODG) method for combined assignment of NMR spectra and 3D structure calculation. The SECODG method implemented in the DIAMOD/NOAH software package dramatically reduces the time to generate protein structures from NMR data. Unlike previous distance geometry methods, which were designed for consistent data sets, the method can generate accurate structures from sets of constraints that contain errors. This project will further improve this method to be applicable for larger proteins. Several tests have demonstrated that the present method can deal with real data and is significantly faster than manual interactive spectral interpretation methods. The first test was to compare the results of automatic structure calculation from NOESY spectra that had been analyzed by earlier manual methods for 6 proteins ranging in size from 40 to 135 amino acids. The automated method assigned 70-80% of the NOESY cross peaks and the three-dimensional structures were of similar quality. In a second test, previously uninterpreted NOESY and TOCSY spectra were used by NOAH/DIAMOD to automatically calculate a 3D structure bundle for an isoform of crambin. Slight modifications and some manual assistance were required so that the program could deal with a significant number of missing chemical shifts. Based on this practical experience, the method will be improved by optimizing the error-tolerant target function, incorporating sensitivity tests for constraints, including line shape information in the peak assignment method and interfacing the program suite with programs for automated sequential assignment from other groups to develop a completely automatic computational package for direct interpretation of NMR spectra. This program package will be a powerful computational tool to speed up macromolecular structure determination from NMR da ta. Experimentally determined three-dimensional structures are the basis for designing new drugs and proteins with improved or new functions. In combination with energy minimization and Monte Carlo simulations, it will help in designing proteins with a desired structure and new functional properties.