Annotating 12 Drosophila genomes for reg- ulatory signals using multiple neural networks and evolutionary history

The application of methods of machine learning is popular in Bioinformatics. Many problems in Bioinformatics can be regarded as pattern recognition problems and well-known methodologies can be used. Until now those methodologies concentrated on the use of a single genome. With the availability of datasets that contain genomes from several closely related species, information that was derived from the evolutionary history of those species can now be utilised in machine learning methods as well. This project will introduce the well-know paradigm of multi classifier systems and deploy them for the recognition of promoter region in the genome of the fruit fly Drosophila. These systems consist of several artificial neural networks whose classification results are merged into a combined statement. The traditional multi classifier system will be extended by incorporating evolutionary parameters. The dataset that is used in this context contains the aligned genomes of 12 different yet closely related species of Drosophila.