Identification of DNA motifs by two different models

Abstract

The identification of DNA regulatory motifs (transcription factor binding sites) in co-regulated genes is essential for understanding the regulatory mechanisms. Many approaches have been developed for motifs discovery in the past. However, identification of regulatory motifs remains a significant challenge. In our opinion, the best motifs-finding methods may be those which rely on exhaustive enumeration because of their high reliability, whereas exhaustive enumeration becomes problematic for long and subtle motifs. In this paper, we present a new method to improve exhaustive enumeration using two different models. We tested its performance on both synthetic and realistic biological data. It proved to be successful in identifying very subtle motifs. Experiments also showed our method outperformed some popular methods in terms of our experimental data.