Tech

Machine learning solves the problem of who is who in the NMR spectrum of organic crystals.

Probabilistic assignment of the 13C NMR spectrum of crystalline strychnine. Credit: @EPFL Manuel Cordova

Using solid-state nuclear magnetic resonance (NMR) spectroscopy, a technique that measures the frequency emitted by the nuclei of several atoms exposed to radio waves in a strong magnetic field, of chemical and 3D structures, and molecules. You can determine the dynamics. And materials.


The first step required for analysis is the so-called chemical shift assignment. This involves assigning each peak of the NMR spectrum to a specific atom of the molecule or material under investigation. This can be a particularly complicated task. Assigning chemical shifts experimentally can be difficult and generally requires time-consuming multidimensional correlation experiments.Allocation by comparison with statistical analysis of experimental chemical shift databases is an alternative solution, but there is no such thing. Database For molecular solids.

A team of researchers including EPFL professors Lindon Msley (Head of Magnetic Resonance Research Institute), Michele Seriotti (Head of Computational Science and Modeling Research Institute) and PhD. Student Manuel Cordova decided to tackle this problem by developing a method for probabilistically assigning the NMR spectra of organic crystals directly from the 2D chemistry.

They began by combining it with the Cambridge Structural Database (CSD), a database of over 200,000 three-dimensional organic structures, to create their own database of chemical shifts in organic solids. ShiftML, The machine learning algorithms they previously developed together can directly predict chemical shifts. structure Of molecular solid.

at first Nature Communications In the 2018 paper, ShiftML uses DFT calculations for training, but it can make accurate predictions for new structures without performing additional quantum calculations. Although DFT accuracy is achieved, this method can calculate chemical shifts in approximately 100 structures. atom In seconds, reduce calculation costs by a factor of 10,000 compared to current DFT chemical shift calculations. The accuracy of this method does not depend on the size of the structure examined and the predicted time is linear with respect to the number of atoms. This sets the stage for calculating chemical shifts in situations that would otherwise be infeasible.

New Science Advances In the paper, the team used ShiftML to predict shifts in over 200,000 compounds extracted from the CSD and associated the resulting shifts with the topology representation of the molecular environment. This is Covalent bond Extends a given number of bonds between atoms within a molecule, away from the central atom. We then put together all the same instances of the graph in the database so that we could get a statistical distribution of the chemical shifts for each motif. This representation is a simplification of the covalent bonds around the atoms in the molecule and does not include the features of the 3D structure. This allows the stochastic assignment of the NMR spectrum of an organic crystal to be obtained directly from the two-dimensional chemical structure. A marginalization scheme that combines distributions from all atoms in a molecule.

After building the chemical shift database, scientists considered predicting allocations in the model system and applied the approach to a set of organic molecules containing carbon. Chemistry Shift assignments have already been experimentally determined, at least in part: theophylline, thymol, cocaine, strychnine, AZD5718, lisinopril, ritonavir, and penicillin G K salt. The allocation probabilities obtained directly from the two-dimensional representation of the numerator are: In most cases, it was found to be consistent with the experimentally determined assignment.

Finally, they evaluated the performance of the framework with a benchmark set of 100 crystal structures with 10 to 20 different carbon atoms. They used the ShiftML predicted shifts for each atom as the correct assignments and excluded them from the statistical distribution used for the assignments. molecule.. The correct assignment was found among the two most likely assignments in over 80% of cases.

“This method has the potential to significantly accelerate the study of materials by NMR by streamlining one of the key first steps in these studies,” says Cordova.


AI and NMR spectroscopy determine atomic configurations at record speeds


For more information:
Manuel Cordova et al, Bayesian Stochastic Assignment of Chemical Shifts in Organic Solids, Science Advances (2021). DOI: 10.1126 / sciadv.abk2341.. www.science.org/doi/10.1126/sciadv.abk2341

Quote: Machine learning is an organic crystal obtained from https://phys.org/news/2021-11-machine-problem-nmr-spectra-crystals.html on November 26, 2021 (November 26, 2021). ) Solves the problem of NMR spectrum.

This document is subject to copyright. No part may be reproduced without written permission, except for fair transactions for personal investigation or research purposes. Content is provided for informational purposes only.



Machine learning solves the problem of who is who in the NMR spectrum of organic crystals.

Source link Machine learning solves the problem of who is who in the NMR spectrum of organic crystals.

Show More

Related Articles

Back to top button