Artificial intelligence (AI) and deep learning are now established as some of the most important technologies of our time. In magnetic resonance imaging (MRI), AI methods are deployed extensively for tasks such as image enhancement and classification, whereas the uptake of AI in nuclear magnetic resonance (NMR) spectroscopy has been slower – but this picture is now swiftly changing.
It will be shown how deep neural networks (DNNs) can be trained for homonuclear decoupling in protein NMR spectroscopy involving the detection of 13C nuclei. 13C-detected methods can be advantageous because they offer superior resolution, however, the spectra are often complicated by homonuclear scalar couplings, which reduce the sensitivity and resolution. Our recent work shows that decoupling of 13C-detected spectra can be achieved by passing a single spectrum through a DNN to yield a singlet spectrum of high quality.
Another area, where we have focussed our developments is for autonomous analysis of complex NMR data. Many NMR tools have been developed to characterise dynamic and exchanging systems; however, analyses of the resulting NMR data often hinge on complex least-squares fitting procedures and human intuition. Deep neural networks will be presented for the analysis of complex 1H chemical exchange saturation transfer (CEST) data, where the DNN not only accurately predicts the chemical shifts of nuclei in the exchanging species, but it also determines the uncertainties associated with these predictions.
All the DNNs developed do not contain any parameters for the end-user to adjust and the methods therefore allows for autonomous analysis of complex NMR data.
References:
1) G Karunanithy, DF Hansen (2021), “FID-Net: A versatile deep neural network architecture for NMR spectral reconstruction and virtual decoupling”, J. Biomol. NMR https://doi.org/10.1007/s10858-021-00366-w
2) G Karunanithy, HW Mackenzie, DF Hansen (2021), “Virtual homonuclear decoupling in direct detection nuclear magnetic resonance experiments using deep neural networks”, J. Am. Chem. Soc https://doi.org/10.1021/jacs.1c04010
3) G Karunanithy, T Yuwen, LE Kay, DF Hansen (2022), “Towards autonomous analysis of chemical exchange saturation transfer experiments using deep neural networks” J. Biomol. NMR https://doi.org/10.1007/s10858-022-00395-z