Genetic algorithms and extraction of rules for detection of short DNA motifs

The paper presents a method for discovery of speciÞc types of rules related to detection and extraction of explicit potentially biologically active DNA motifs from nucleotide databases. The characteristic of these rules is that they represent a relation of the strengths of signals of two motifs and their mutual distance. The rule extraction is based on a genetic algorithm. The method is applied and tested in the extraction of explicit rules that govern the relationship of the TATA-box motifs in eukaryotes, the signal that relates to the [−40,+11] region relative to the transcription start site (TSS) of eukaryotic promoters, and the distance of the TATA motif and TSS. A very good discrimination ability of the extracted rules in separation of the ’presumed biologically functional’ TATA motifs and de-facto non-functional (pseudo) TATA motifs is demonstrated.