LOGOS: a modular Bayesian model for de novo motif detection

E.P. Xing, W. Wu, M.I. Jordan, R.M. Karp
Computational Systems Bioinformatics. CSB2003. Proceedings of the 2003 IEEE Bioinformatics Conference. CSB2003  
127 128 E. P. Xing et al. superior performance on both semi-realistic test data and cis-regulatory sequences from yeast and Drosophila genomes with regard to sensitivity, specificity, flexibility and extensibility. a Not to be confused with model-based motif scan, the task of searching known motifs based on given position weight matrices, as addressed by Frith et al. 10 and Huang et al. 15 c Heuristics are generally employed -such as throwing away overlapping sampled motifs (in the Gibbs
more » ... ) or rescaling the joint posterior of x (in MEME) -to enforce the non-overlapping constraint. Nevertheless, this results in inconsistencies between the computed motif distribution and the one defined by the model, and incurs a sizable overhead due to wasteful computations.
doi:10.1109/csb.2003.1227327 dblp:conf/csb/XingWJK03 fatcat:f37iwhtedndznjqn4yfl7lhymi