Mining the organic compound jungle-a functional programming approach

Pharmaceutical companies are continually striving to determine the common key characteristics of compounds that determine their functionality (e.g. relief of asthma) so that they may continue to provide safe and effective medicines. Historically this has been carried out by visually comparing graphical representations of the structures of compounds, which possess the same functionality, so that key substructures (pharmacophores) may be determined. However, with the advent of high throughput screening techniques providing data on enormous numbers of compounds, this has become inappropriate. (A human can only compare a certain number of patterns accurately in a day.) Potential solutions to this problem may appear to come from a knowledge based systems approach based upon pattern matching. However, we suggest this solution does lie with a knowledge based systems approach but one which relies on data mining as the underpinning technology. We discuss our initial work which shows that organic compounds, possessing the same functionality, may be mined for common substructures using data mining techniques. We also discuss how our prototype tool, in which a selection of data mining algorithms may be chosen, has been developed in a functional programming language. The functional language Gofer, which was used to rapidly prototype the tool, readily lends itself to the task due to its polymorphism and lazy evaluation. The lazy evaluation of Gofer is a particularly useful feature of the language which readily enables the common characteristics to be determined no matter how large the compounds.