MProfiler: A Profile-Based Method for DNA Motif Discovery

Motif Finding is one of the important tasks in gene regulation which is essential in understanding biological cell functions. Based on Tompa et al. study, the performance of current motif finders is not satisfactory. A number of ensemble methods has been proposed to enhance the results. Existing ensemble methods overall performance is better than stand-alone motif finders. A recent ensemble method, MotifVoter, significantly outperforms all existing stand-alone and ensemble methods. In this paper we propose a method, MProfiler, to increase the accuracy of MotifVoter without increasing the running time by introducing an idea called center profiling. Our test shows improvement in the quality of generated clusters over MotifVoter in both accuracy and cluster compactness. Using 56 datasets, the accuracy of the final results using our method has 80% improvement in correlation coefficient nCC, and 93% improvement in performance coefficient nPC over MotifVoter.