A combined NMR and Deep Neural Network approach for enhancing the spectral resolution of aromatic side chains in proteins

Nuclear magnetic resonance (NMR) spectroscopy has become an important technique in structural biology for characterising the structure, dynamics and interactions of macromolecules. While a plethora of NMR methods are now available to inform on backbone and methyl-bearing side-chains of proteins, a characterisation of aromatic side chains is more challenging and often requires specific labelling or 13C-detection. Here we present a deep neural network (DNN) named FID-Net-2, which transforms NMR spectra recorded on simple uniformly 13C labelled samples to yield high-quality 1H-13C correlation spectra of the aromatic side chains. Key to the success of the DNN is the design of a complementary set of NMR experiments that produce spectra with unique features to aid the DNN produce high-resolution aromatic 1H-13C correlation spectra with accurate intensities. The reconstructed spectra can be used for quantitative purposes as FID-Net-2 predicts uncertainties in the resulting spectra. We have validated the new methodology experimentally on protein samples ranging from 7 to 40 kDa in size. We demonstrate that the method can accurately reconstruct high resolution two-dimensional aromatic 1H-13C correlation maps, high resolution three-dimensional aromatic-methyl NOESY spectra to facilitate aromatic 1H-13C assignments, and that the intensities of peaks from the reconstructed aromatic 1H-13C correlation maps can be used to quantitatively characterise the kinetics of protein folding. More generally, we believe that this strategy of devising new NMR experiments specifically for analysis using customised DNNs represents a substantial advance that will have a major impact on the study of molecules using NMR in the years to come.核磁共振波谱法（NMR spectroscopy）现已成为结构生物学中表征大分子结构、动态特性与相互作用的重要技术手段。尽管目前已有大量NMR方法可用于解析蛋白质主链与含甲基侧链的相关信息，但对芳香族侧链的表征仍颇具挑战，通常需要采用特定的同位素标记或13C检测技术。本研究提出一款名为FID-Net-2的深度神经网络（DNN），可对简单均匀13C标记样品采集的NMR谱图进行转换，从而得到高质量的芳香族侧链1H-13C相关谱。该神经网络得以成功应用的核心在于，我们设计了一套互补的NMR实验方案，所生成的谱图具备独特特征，可辅助神经网络生成具有准确峰强的高分辨率芳香族1H-13C相关谱。由于FID-Net-2可预测生成谱图的不确定度，因此重构后的谱图可用于定量分析。我们已针对分子量7至40 kDa的蛋白质样品，通过实验验证了该新方法的有效性。本研究证明，该方法可精准重构高分辨率二维芳香族1H-13C相关谱图、高分辨率三维芳香族-甲基NOESY谱，以辅助芳香族1H-13C信号归属，且重构得到的芳香族1H-13C相关谱图的峰强可用于定量表征蛋白质折叠的动力学过程。总体而言，我们认为这种专为定制化深度神经网络分析而设计新型NMR实验的策略，是一项重大进展，未来将对基于NMR的分子研究产生深远影响。