Molecular structure descriptors in the computer-aided design of biologically active compounds

The current state of description of molecular structure in computer-aided molecular design of biologically active compounds by means of descriptors is analysed. The information contents of descriptors increases in the following sequence: element-level descriptors–structural formulae descriptors–electronic structure descriptors–molecular shape descriptors–intermolecular interaction descriptors. Each subsequent class of descriptors normally covers information contained in the previous-level ones. It is emphasised that it is practically impossible to describe all the features of a molecular structure in terms of any single class of descriptors. It is recommended to optimise the number of descriptors used by means of appropriate statistical procedures and characteristics of structure–property models based on these descriptors. The bibliography includes 371 references.