Bisphosphine ligand conformer selection to enhance descriptor database representation: improving statistical modelling outcomes

A foundational consideration in the development of computationally derived molecular feature libraries is the generation and selection of conformers. It has been shown that several feature values have a degree of conformer depencency – which may have significant mechanistic implications, partiticulary in the field of homogeneous enantioselective catalysis. However, the computational cost of calculating conformers often prohibits this analysis from being performed, especially when large flexible systems are involved. We report here a practical, chemically-intutive conformer selection tool for bisphosphine-ligated palladium(ii) dichloride complexes that provide a good balance between representation and computational cost. Conformer-weighted features generated from this method were applied to two previous statistical modelling case studies, where weighted features improve model quality with respect to predictive power. This selection methodology has the potential to be applied to a range of complex molecular systems beyond bisphosphine-ligated organometallic complexes.