DEEP Phaser: A Deep Learning Tandem Vision Transformer for Fully Automated NMR Phase Correction

Although phase correction is one of the most routine steps in NMR data processing, even the best available automated approaches often require manual adjustments by human experts. A deep learning-based phase correction algorithm is presented as a tandem vision transformer artificial neural network. It has been trained on a large set of synthetic solution-NMR like spectra and determines the zeroth- and first-order phase correction based on the entire input spectrum and achieves very high phasing accuracy for a broad range of experimental spectra without requiring any further manual adjustments. The new method, called DEEP Phaser, is demonstrated for a variety of different real-world solution ¹H 1D NMR spectra, from small molecules and their complex mixtures to biomacromolecules, and is available as free software and as a public web server.