BBQ in Tanimoto Scores

After the sequencing of the complete genomes of several higher organisms, molecular biologists face new challenges in understanding the regulatory mechanisms which control the expression of protein coding genes. The expression of a gene into a functional product is a fairly complex process consisting of several steps, with each step underlying some regulatory mechanism. The first major step in expression of protein coding genes is the transcription of the DNA into a messenger-RNA by the RNA polymerase II. For initiation of transcription the polymerase must locate and bind to a site upstream of the protein-coding region of the gene. This binding is controlled by several proteins, known as transcription factors, which also bind to the DNA and stimulate or repress the formation of the initiation complex around the RNA polymerase. DNA binding sites of transcription factors are known to occur in clusters among a small range, commonly referred to as cis-regulatory modules. Since transcription factors affect transcription control, genes which share the same factors or modules tend to have the same expression patterns – they are said to be co-expressed. In this work we deal with algorithms for discovering such cis-regulatory modules, which occur in the regulatory regions of multiple different genes. The presented Best-Barbecue-Problem provides the formal algorithmic foundation of a discovery method, which searches in a set of genomic sequences for the largest clusters of binding sites, that are located within a window of a certain length and occur in all of the input sequences. The bbq program solves this NP-complete optimization problem. Although bbq always finds the best solution, it fails in certain instances of input data. If only some of the sequences do not share a few binding sites with the others, the overall result is restricted to the lowest common denominator. The topic of this thesis is to overcome these limitations by extending the Best-Barbecue algorithm by novel scoring schemes. The search is now aimed on finding sets of binding sites with largest possible cardinality which occur in the majority of the sequences. We implemented three scoring functions based on the so-called Tanimoto score and integrated them into the bbq algorithm. To evaluate the performance of the new search method, we tested bbq with two biologically relevant sets of regulatory sequences which have been studied before.