Estimation of Melting Temperature of Molecular Cocrystals Using Artificial Neural Network Model

A quantitative structure–activity relationship model has been constructed by artificial neural networks for estimation of melting temperature ( T m ) of molecular cocrystals (CCs). On the basis of a literature analysis using SciFinder and Cambridge Structural Database softwares, a database has been created of CCs for four active pharmaceutical ingredients, namely, caffeine, theophylline (THP), nicotinamide (NA), and isonicotinamide (INA). In total, of 61 CCs were included: 14-CAF, 9-THP, 29-INA, and 9-NA. A good correlation was obtained with ANNs to quantify the T m of the CCs with respect to various coformers. The training process was completed with an average relative error of 2.38%, whereas the relative error for the validation set was 2.89%.