IMPRESSION generation 2 – accurate, fast and generalised neural network model for predicting NMR parameters in place of DFT.
Year: 2025
Type: article
Abstract: scalar coupling constants) and, critically, it also demonstrates generalisability when tested against molecules from sources that are completely independent of its own training data. When compared to experimental NMR data for ∼5000 compounds, IMPRESSION-G2 gives results in minutes on a standard laptop which are almost indistinguishable from DFT results that took days on a large scale High Performance Computing system. This accuracy and speed of IMPRESSION-G2 coupled to GFN-xTB shows that it can be used to simply replace DFT for predicting 3D-aware NMR parameters inside the wide chemical space of its training data.
Source: Chemical Science
DOI: 10.1039/d4sc07858f
Institutions: University of Bristol, Genentech, Evotec (United Kingdom)
Language: English
Subfield: Spectroscopy
Field: Chemistry
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4408993583