Nature Computational Science has published a highlight of the article "Deep Learning Metal Complex Properties with Natural Quantum Graphs" recently published in Digital Discovery by Hannes Kneiding, Ruslan Lukin, Lucas Lang, Simen Reine, Thomas Bondo Pedersen, Riccardo de Bin, and David Balcells.
The highlight mentions the main features of this work, including the modeling of complex bonding patterns between transition metals and organic ligands, the use of NBO analysis data for building and informing graph representations for transition metal complexes, the prediction of quantum properties with graph neural networks that leverage both geometric and electronic structure information, and the computation of an open dataset containing 60k graphs.
This work is part of the PhD studies of Hannes Kneiding, which is supported by the COMPSCI project funded by the EU MSCA-COFUND program, and it is also a key part of the FRIPRO catLEGOS project led by David Balcells and funded by the Research Council of Norway.