COMPAS-4: a Dataset of (BN)- Substituted Cata-Condensed Polybenzenoid Hydrocarbons – Data Analysis and Feature Engineering

Incorporation of a BN pair into polycyclic aro- matic hydrocarbons is a common approach for modulating their electronic properties. However, a conceptual and quantitative framework rationalizing the observed effects has not been developed, and general structure-property relationships remain elusive. In this work, we perform a data-driven investigation that leads to concrete principles for rational design of (BN)1-PBHs with targeted properties. We construct a new chemical database, COMPAS-4, which contains the geometries and properties of all possible (BN)1-PBH isomers up to 6 rings, calculated at both the GFN1-xTB and DFT (CAM-B3LYP/def2-SVP) levels of theory. We investigate the influence of BN-substitution on various molecular properties, including their molecular orbital energies and aromaticity, and define specific structural features that determine these properties. Notably, all of these features are chemically intuitive and simple to extract from the structure of the molecule, without any prior computation. We find that the most influential feature is the number of rings whose cyclic delocalization is disturbed as a result of the substitution