A combinatorial exploration of a very large set of azaboles

The thermodynamic properties of 778 azaboles, including the known pyrazaboles, imidazaboles, and triazaboles, have been theoretically calculated at the CBS-QB3 level. Pólya enumeration theorem was used to determine the number of skeletons of azaboles (237 neutral systems), followed by an exploration and generation of the whole space of N–H tautomers. The experimental X-ray geometries, B–C and B–N distances and folding angles, were compared with those of the calculated geometries. Using Free-Wilson (or absence-presence) matrices and multiple regression analysis, the fragments’ contribution to different enthalpies was calculated. Finally, a machine learning model based on Gaussian Process Regression has been created to predict the ΔH°_f of the azaboles together with their structural characteristics.