Boolean genetic programming for promoter recognition in eukaryotes

Fixed-length genetic programming is applied to the problem of promoter identification in eukaryotes. The goal is to generate solutions that can be easily interpreted and compared with known promoter characteristics. Using a Boolean function set applied to Boolean registers, inputs, and constant values, the approach builds a logical expression whose value gives the classification decision. Evaluated on a dataset of human promoters and non-promoters from coding regions, the approach is found to generate concise solutions that yield good specificity but poor sensitivity. Analysis of the programs that are generated indicates that a well-known, biologically significant, characteristic of promoter regions is successfully identified. Suggested future work involves implementing the system using fuzzy logic