The aim of the study was to investigate whether easily and rapidly calculated 2D and 3D molecular descriptors could predict the melting point of drug-like compounds, to allow a melting point classification of solid drugs. The melting points for 277 structurally diverse model drugs were extracted from the 12th edition of the Merck Index. 2D descriptors mainly representing electrotopology and electron accessibilities were calculated by Molconn-Z and the AstraZeneca in-house program Selma. 3D descriptors for molecular surface areas were generated using the programs MacroModel and Marea. Correlations between the calculated descriptors and the melting point values were established with partial least squares projection to latent structures (PLS) using training and test sets. Three different descriptor matrixes were studied, and the models obtained were used for consensus modeling. The calculated properties were shown to explain 63% of the melting point. Descriptors for hydrophilicity, polarity, partial atom charge, and molecular rigidity were found to be positively correlated with melting point, whereas nonpolar atoms and high flexibility within the molecule were negatively correlated to this solid-state characteristic. Moreover, the studied descriptors were successful in providing a qualitative ranking of compounds into classes displaying a low, intermediate, or high melting point. Finally, a mechanism for the relation between the molecular descriptors and their effect on the melting point and the aqueous solubility was proposed.