| In recent years, along with the rapid advances in sequencing technology in biology, there is also a great breakthrough in the progress of the research in modern biological technology. Bio-sequence data grows at an unprecedented speed. Manual analysis and processing of biological sequence data has been unable to meet the needs. How to analyze and deal with such a large amount of data, as well as understand the biological significance of this data have become the most important research tasks. Motif discovery in DNA sequences is a fundamental problem in bioinformatics, which relates to many biological problems such as gene finding, transcription factor binding site finding and promoter finding. At present, biologists have turned their eyes on multiple genes association from single gene which takes role in lives'character. Motifs in regulate region regulate play an important role in gene transfer, so it is of great significance to study these motifs. At the same time identification of these motifs is important to decode genome.Many algorithms have been developed to tackle motif discovery problem, e.g. Gibbs, MEME and so on. In this paper, we firstly analyzed the models used in motif discovery algorithms. Also, we studied the motif discovery algorithms mentioned above based on different models. It is assumed that the positions within a motif are mutually independent in these traditional algorithms. Recent biological experiments, however, suggest that positions within a motif are not completely independent and there exist interdependency among positions in some motifs. This interdependency should be exploited to improve motif discovery. Based on this consideration, we introduce interdependency into scoring function and apply it to Gibbs sampling to develop an algorithm PIGS (Position Interdependency based Gibbs Sampler) for motif discovery. The simulated results on the synthetic and real data show great improvement in accuracy and sensitivity of finding true motifs. |
PDF Full Text Request