High-throughput computational workflow for ligand discovery in catalysis with the CSD

Abstract

A novel semi-automated, high throughput computational workflow for ligand/catalyst discovery based on the Cambridge Structural Database is reported. The transition states of the rate-determining step of the Ullmann-Goldberg reaction were...

Automated Transition Metal Catalysts Discovery and Optimisation with AI and Machine Learning

“…Nguyen group adopted a different approach to leverage the wide chemical space covered by the Cambridge Structural Database (CSD), using both organic and organometallic structures, and to avoid the question of synthetic viability of the ligands. [51] Two closed-shell TS of the rate-determining-step (RDS) of the Ullmann-Goldberg coupling reactions were optimised with DFT (DLPNO-CCSD(T)/def2-TZVP). These TS structures were used as template to create the catalophores with the desired geometry and empty space for the Cu(I) cation and the substrates.…”

“…[52,53] Modifications has been made to MolSimplify to enable it to build TS with asymmetrical geometry and unusual coordination numbers. [51] Alternatively, the CSD Python tool is also a highly flexible tool and for our own purposes used it to build organometallic complexes. [54] The CSD python API loads the molecule as a class object and can build and edit molecules, such as adding, removing bonds and atoms, and normalising charges and hydrogens.…”

“…Usually, lower level molecular modelling methods are employed to generate their training and validation data, after benchmarking against higher level DFT methods to establish their accuracy. This is even more challenging when TS properties are used as training data, [51] as optimisation of TS demands much more CPU time than optimisation of stable intermediates, ligands and starting materials. Furthermore, optimisation of TS is also more prone to errors and failures, which can lead to large amount of unproductive CPU time.…”

Large Computational Survey of Intrinsic Reactivity of Aromatic Carbon Atoms with Respect to a Model Aldehyde Oxidase

“…While it has become quite customary to report generation of thousands of TSs in a single research project, − such efforts should be encouraged. The accumulation of TS data portends a future qualitative transition in our ability to model and predict arbitrary chemical reaction pathways with machine learning methods, which require large amounts of training data. ,,, …”

“…The accumulation of TS data portends a future qualitative transition in our ability to model and predict arbitrary chemical reaction pathways with machine learning methods, which require large amounts of training data. 28,77,79,80 All of the TS barriers that were obtained as a result of application of our automated workflow to the set of heterocycles are presented in the Supporting Information. Here, we draw conclusions about the workflow performance, discuss reactivity trends observed upon the analysis of the outcome of the calculations, and point out caveats.…”