基于压缩感知与重采样技术的NMR噪声抑制新方法

Developing high sensitive measurement method is an everlasting topic in analytical chemistry. The essential approach of sensitivity enhancement is enhancing signal, or suppressing noise. In nuclear magnetic resonance spectroscopy (NMR), it is usual to enhance signal by utilizing high field magnitude spectrometer or complicated pulse sequences, and to suppress noise by equiping cryogenic probe, which consequently increases experimental cost and difficulty. By comparison, data post-processing is more cost-effective, and has wider developing prospective. Previously, we proposed a data processing method based on statistic resampling principle to suppress noise and artifact in NMR spectra, which has been verified. Starting from the above idea, this project will introduce compressed sensing, and combine it with resampling, to develop novel noise suppression methods. The modular general data processing software package which is capalbe to deal with NMR data in typical manufactures'spectrometers or workstations also will be developed. All the developed new methods will be applied in fast multidimensional biological NMR study, especially the NMR structure study of low soluble protein, to increase NMR spectral equality and elevate the efficiency of resolving protein structure. The project result also may be applicable to data processing of other spectroscopies base on Fourier transform.

发展高灵敏检测方法是分析化学的永恒主题之一，提高信号强度和降低噪声水平是增强灵敏度的根本途径。在核磁共振波谱（NMR）分析中，通常采用高磁场强度的谱仪或复杂的脉冲实验方法来提高信号强度，通过使用超低温探头来降低噪声水平，但这无疑会提高实验成本或增加实验难度。相较而言，利用数据后处理方法辨识和抑制噪声，是提高信噪比的一种更为经济的途径。我们前期发展了基于统计学中重采样原理、能够同时抑制NMR谱中噪声/伪峰的数据后处理方法，并获得初步验证。以此为基础，本项目拟进一步引入压缩感知技术，将其与重采样方法相结合，发展基于数据后处理的新型噪声抑制方法，开发可直接用于不同厂家NMR谱仪/工作站的模块化普适性数据处理软件，并将其应用到快速多维生物NMR研究（特别是低可溶性蛋白质结构的NMR研究）中，以期提高NMR谱图质量，加快蛋白质结构解析效率。本成果预期也适用于其它基于傅立叶变换的谱学数据处理。

核磁共振波谱作为一种谱学分析工具，在化学分析领域有着重要应用，近年来更发展成为结构生物学、代谢组学等热点学科中不可或缺的分析手段。然而与其它谱学分析方法如质谱等相比，核磁共振波谱的灵敏度较弱，这使得核磁共振检测需要更多的样品量，或者更长的采样时间，才能获得足够信噪比的谱图。因此，当样品量很少，或者样品不够稳定时，核磁共振检测就遇到困难。在核磁共振波谱实验中，通常采用高场谱仪或复杂的脉冲实验方法来提高信号强度，通过超低温探头来降低噪声水平，但这无疑会提高实验难度或成本。相较而言，利用数据后处理方法辨识和抑制噪声，是提高信噪比的一种更为经济的途径。我们发展了几种基于统计重采样和压缩感知谱图重建的数据处理方法，可以用于噪声、采样截断伪峰、t1噪声的有效抑制，从而提高NMR谱图信噪比。