Encoded molecular diversity methods

JJ Baldwin, RE Dolle. Deconvolution methods in solid-phase synthesis. In WH Moos, MR Pavia, BK Kay, AD Ellington, eds. Annual Reports in Combinatorial Chemistry and Molecular Diversity. Leiden ESCOM, 1997, pp. 287-297 X-Y Xiao, MP Nova. Radiofrequency encoding and additional techniques for the structure elucidation of synthetic combinatorial libraries. In SR Wilson, AW Czarnik, eds. Combinatorial Chemistry Synthesis and Application. New York Wiley, 1997, pp. 135-152. 

Conceptually, the problem of quantifying molecular diversity involves two parts the first is the definition of chemical distance, and the second is the selection of a representative set of compounds from a (typically) much larger collection. This section reviews a number of methods that have been proposed to encode molecular structures in a way that is suitable for numerical processing. It is organized in three parts, which discuss descriptors derived from the topology, three-dimensional structure, and physicochemical and electronic properties of the molecules. 

The surface tension of a liquid is a measure of its tendency to minimize its surface area. The models for surface tension found in the literature are built on thermodynamic approaches, and they relate surface tension to a number of other physical properties, or to combinations of them. However, the literature contains little concerning the effect of specific molecular features on surface tension and provides no method to calculate surface tension from molecular structure. A set of 146 values for surface tension at 30 °C was extracted from a paper by Jasper that reports values for more than 2200 pure compounds with diverse structures, often at several different temperatures and with an experimental error of approximately 0.10 dyn cm . The compounds were encoded using a variety of topological, geometric, and electronic descriptors. A model was developed for a combined set of alkanes, alkyl esters, and alkyl alcohols which utilized 10 descriptors and had s = 0.4 dyn cm (1.8% of the mean). This model was then used to predict the surface tensions for 20 compounds not used in model development. 

---