Towards autonomous analysis of chemical exchange saturation transfer experiments using deep neural networks

Abstract

Macromolecules often exchange between functional states on timescales that can be accessed with NMR spectroscopy and many NMR tools have been developed to characterise the kinetics and thermodynamics of the exchange processes, as well as the structure of the conformers that are involved. However, analysis of the NMR data that report on exchanging macromolecules often hinges on complex least-squares fitting procedures as well as human experience and intuition, which, in some cases, limits the widespread use of the methods. The applications of deep neural networks (DNNs) and artificial intelligence have increased significantly in the sciences, and recently, specifically, within the field of biomolecular NMR, where DNNs are now available for tasks such as the reconstruction of sparsely sampled spectra, peak picking, and virtual decoupling. Here we present a DNN for the analysis of chemical exchange saturation transfer (CEST) data reporting on two- or three-site chemical exchange involving sparse state lifetimes of between approximately 3–60 ms, the range most frequently observed via experiment. The work presented here focuses on the 1H CEST class of methods that are further complicated, in relation to applications to other nuclei, by anti-phase features. The developed DNNs accurately predict the chemical shifts of nuclei in the exchanging species directly from anti-phase 1HN CEST profiles, along with an uncertainty associated with the predictions. The performance of the DNN was quantitatively assessed using both synthetic and experimental anti-phase CEST profiles. The assessments show that the DNN accurately determines chemical shifts and their associated uncertainties. The DNNs developed here do not contain any parameters for the end-user to adjust and the method therefore allows for autonomous analysis of complex NMR data that report on conformational exchange.

An introduction to one- and two-dimensional lineshape analysis of chemically exchanging systems

“…Lineshape fitting methods can also be applied to parametric estimation problems separate from chemical exchange, for example in the analysis of methyl side-chain dynamics by fitting 1 H-coupled 13 C multiplets to determine cross-correlated relaxation rates [79]. There is the potential for these methods to be coupled with automated, optimal acquisition protocols [80], and for the application of machine learning methods to accelerate the interpretation of NMR observations [81]. Thus, we are confident that this technique has a bright future that will find increasing applications across a variety of chemical and biological systems.…”

Solution State Methyl NMR Spectroscopy of Large Non-Deuterated Proteins Enabled by Deep Neural Networks

“…As described previously, the output from each layer is combined to give the final output and also added to the input for the residual unit to form the input for the next layer. For the 13 C network, the dilations employed are cycled through the values: 1,2,4, 6,8,10,12,14,16,20,24,28,32,40,48,56,64, and there are 128 filters for each convolutional layer. For the 1 H network the dilations employed are 1,2,4,6,8,10,12,14,16,20,24,28,32, and there are 64 filters per convolutional layer.…”

“…For the 13 C network, the dilations employed are cycled through the values: 1,2,4, 6,8,10,12,14,16,20,24,28,32,40,48,56,64, and there are 128 filters for each convolutional layer. For the 1 H network the dilations employed are 1,2,4,6,8,10,12,14,16,20,24,28,32, and there are 64 filters per convolutional layer.…”

“…In this current era with burgeoning applications and developments of AI, from computational structural biology 4,5 to sophisticated large language models 6 , it is natural to look for solutions within the field of AI to the challenges encountered in characterising large proteins. In this context, it has over the last couple of years been shown by others and us that deep neural networks (DNNs) can indeed be trained to accurately transform [7][8][9] and analyse 10,11 complex NMR data. The most recent applications use supervised deep learning, where a DNN is supplied with an input and target training dataset and through a training process the DNN 'learns' the mapping between the two.…”

Solution-state methyl NMR spectroscopy of large non-deuterated proteins enabled by deep neural networks

“…In this era of burgeoning applications and developments in AI, from computational structural biology 4,5 to sophisticated large language models 6 , it is natural to look for solutions within this field for the challenges encountered in characterising large proteins. In this context, we and others have recently demonstrated that deep neural networks (DNNs) can be trained to accurately transform [7][8][9] and analyse [10][11][12] complex NMR data. The most recent applications use supervised deep learning, where a DNN is supplied with an input and a target training dataset and through a training process the DNN attempts to determine the mapping between the two.…”