Stereoelectronics-Aware Molecular Representation Learning

Abstract

The representation of molecular structures is crucial for molecular machine learning strategies. Although graph representations are highly versatile and show their broad applicability, they lack information about the quantum-chemical properties of molecular structures. This work proposes a new way to infuse such information into molecular graphs, using a supervised learning method. As a result, the model is able to predict essential higher-order interactions between electron-rich and electron-deficient localized orbitals. The learned interactions are then used as a representation for the prediction of downstream tasks, improving over QM9 baselines.

Deep learning metal complex properties with natural quantum graphs

“…Further, the NatQG graphs encode the NBO orbitals implicitly, embedding their defining parameters into the node and edge attribute vectors of a molecular graph that, especially in the case of u-NatQG , can be directly related to the skeletal formula of the represented TMC. An alternative approach, recently explored by Gomes et al , 89 consists in representing LP and BD orbitals with additional explicit nodes. The NatQG graphs may also be used to develop a string representation with rich electronic structure information, similar to the representation developed by Dietz.…”

Deep Learning Metal Complex Properties with Natural Quantum Graphs

“…Further, the NatQG graphs encode the NBO orbitals implicitly, embedding their defining parameters into the node and edge attribute vectors of a molecular graph that, especially in the case of u-NatQG, can be directly related to the skeletal formula of the represented TMC. An alternative approach, recently explored by Gomes et al, 84 consists in representing LP and BD orbitals with additional explicit nodes. The NatQG graphs may also be used to develop a string representation with rich electronic structure information, similar to the representation developed by Dietz.…”