Improving motif refinement using hybrid expectation maximization and random projection

Abstract

The main goal of the motif finding problem is to detect novel, over-represented unknown signals in a set of sequences. Popular algorithms like Expectation Maximization (EM) and Gibbs sampling are sensitive to the initial guesses and are known to converge to the nearest local maximum very quickly. A novel optimization framework searches the neighborhood regions of the initial alignments in a systematic manner to explore the multiple local optimal solutions. This effective search is achieved by transforming the original optimization problem into its corresponding dynamical system and estimating the practical stability boundary of the local maximum. The work aims at implementing the hybrid algorithm and enhancing it by trying different global methods and other techniques. Then aggregation methods rather than projection methods are tried.