Towards quantum informed atom pairs

In the following research, a new modification of traditional atom pairs is studied. The atom pairs are enriched with values originating from quantum chemistry calculations. Random forest machine learning algorithm is applied in modelling 10 different properties and biological activities based on different molecular representations and evaluated in repeated cross-validation. The predictive power of modified atom pairs - quantum atom pairs are compared to the predictive powers of traditional molecular representations known and widely applied in cheminformatics. Root mean squared error, $R^2$, the area under the receiver operation curve and balanced accuracy are used to evaluate the predictive power of applied molecular representations. The research shows that while performing regression tasks, the quantum atom pairs provide better fitting to the data than their precursors.