Accelerated Nuclear Magnetic Resonance Spectroscopy with Deep Learning

Abstract

Nuclear magnetic resonance (NMR) spectroscopy serves as an indispensable tool in chemistry and biology but often suffers from long experimental times. We present a proof‐of‐concept of the application of deep learning and neural networks for high‐quality, reliable, and very fast NMR spectra reconstruction from limited experimental data. We show that the neural network training can be achieved using solely synthetic NMR signals, which lifts the prohibiting demand for a large volume of realistic training data usually required for a deep learning approach.

Deep metabolome: Applications of deep learning in metabolomics

“…The model performance was validated with an input of 2D 1 H- 15 N HSQC spectrum with 25% NUS data quality against the fully sampled 2D and 3D spectra and obtained a correlation of peak intensities of 0.99. This model also displayed correlation coefficient greater than 0.98 to 2D spectra [41] even in low-density regions.…”

“…The model was later validated and compared against sparse multidimensional iterative lineshape-enhanced (SMILE), hmsIST algorithms by using the experimental 15 N- 1 H HSQC spectrum. The DL-based approach showed equally good or slightly better NMR spectra reconstruction results compared with current state of the field methods [40] , [41] proposed to use a CNN to reconstruct fast and high-quality NMR spectra of small and large (metabolites) and small proteins from fully simulated NMR data [41] . The model performance was validated with an input of 2D 1 H- 15 N HSQC spectrum with 25% NUS data quality against the fully sampled 2D and 3D spectra and obtained a correlation of peak intensities of 0.99.…”

Clustering NMR: Machine learning assistive rapid two-dimensional relaxometry mapping

“…Machine learning in the form of dimension reductionist (e.g., principal component analysis (PCA), partial least squares (PLS)) have also been used to reduce the dimensionality in multidimensional spectroscopy (e.g., NMR metabolomics [19,35,36] ). A recent deep learning assistive NMR spectroscopy [18] , which signals reconstructing were demonstrated. We summarized and compared Clustering NMR method with the state-of-the-art methodologies in a SWOT-like analysis (Table 2).…”

“…(1675 words) [19] , Karaman (2015) [36] , Rocha (2018) [35] spectroscopy metabolomics 2D PCA/PLA informative slow Frydmann (2014) [30] spectroscopy ultrafast NMR 2D no rapid gradient field Qu (2019) [18] spectroscopy generic ndimensional deep learning speed up information lose? Haun (2010) [40] , Haun [20] (2011), Liong (2013) [37] , Peng (2014) [21] , Neely (2016) [41] , Robinson (2017) [42] relaxometry medical diagnosis 1D no rapid, PoCT low specificity and sensitivity, missing out cross peaks?!…”

“…High resolution nuclear magnetic resonance (NMR) spectroscopy is a powerful and attractive approach in biochemistry (e.g., protein analysis [1] , metabolomics [2][3][4] ) and inorganic chemistry [5] . In the recent years however, with the rapid advances in NMR engineering (e.g., IC-based spectrometer [6][7][8][9][10][11][12] , microfluidic-based chip [13][14][15][16][17] , artificial intelligence [18,19] ) utilizing small foot-print permanent magnet, the time-domain NMR instrumentations have seen a myriad of interesting applications from point-of-care testing (PoCT) medical diagnosis [7,[20][21][22][23] , industrial food science [24,25] , and in-situ oil-gas exploration [26,27] . Biochemical information is typically detected and encoded in the frequency domain ( chemical shift ) in the high-field NMR.…”