De-Novo motif finding using genetic algorithm

One of the central problems in bioinformatics is de-novo finding of recurring motifs in the DNA. Since these motifs are preserved throughout evolution they probably have a significant biological role. One of the most widely used existing tools uses Expectation Maximization (EM) algorithm in order to learn the parameters of a statistical model based on partial data. One such method is based on assuming the Motif-data is generated by a Hidden Markov Model (HMM). This method is called the MEME algorithm. Despite its success, this method is in its essence a hill-climbing method, and as such, is known to be subject to being caught in local optima. In this work, we tackled the problem by using, instead, a genetic algorithm, and to search for the optimal probabilities of the HMM model. In certain occasions we succeeded in achieving better results using GA.