Quantitative Structure Activity Relationship (QSAR) is an approach of mapping chemical structure to properties. A significant development can be observed in the last two decades in this method which originated from the Hansch analysis based on the logP data and Hammett constant towards a growing importance of the molecular descriptors derived from 3D structure including conformational dynamics and solvation scenarios. However, molecular interactions in biological systems are complex phenomena generating extremely noisy data, if simulated in silico. This decides that activity modeling and predictions are a risky business. Molecular recognition uncertainty in traditional receptor independent (RI) m-QSAR cannot be eliminated but by the inclusion of the receptor data. Modeling ligand-receptor interactions is a complex computational problem. This has limited the development of the receptor dependent (RD) m-QSAR. However, a steady increase of computational power has also improved modeling ability in chemoinformatics and novel RD QSAR methods appeared. Following the RI m-QSAR terminology this is usually classified as RD 3÷6D-QSAR. However, a clear systematic m-QSAR classification can be proposed, where dimension m refers to, the static ligand representation (3D), multiple ligand representation (4D), ligand-based virtual or pseudo receptor models (5D), multiple solvation scenarios (6D) and real receptor or target-based receptor model data (7D).