Abstract:
|
Using computational means to help understand gene regulation is one of the major challenges in the post-genome era. In the past few years, Gibbs sampling-based algorithms have been developed for finding novel regulatory binding motifs in DNA sequences. A successful application of these algorithms is to examine the upstream regions of the genes in a cluster, which is determined by the genes' expression profiles. Another application is to conduct a cross-species comparative motif analysis. Main disadvantages of these algorithms, however, like many other Gibbs-sampling algorithms, are that they are easily trapped in a local mode and, thus, take a long time to run. I will report a drastically different, but speedier and more accurate motif finding the algorithm we recently developed. I will also describe a Bayesian motif clustering method for post-processing the predicted motif patterns, which is useful for discovering co-regulated genes.
|