Unsupervised Learning of Multiple Motifs in Biopolymers Using Expectation Maximization
Machine Learning - Special issue on applications in molecular biology
An introduction to variational methods for graphical models
Learning in graphical models
A generalized mean field algorithm for variational inference in exponential families
UAI'03 Proceedings of the Nineteenth conference on Uncertainty in Artificial Intelligence
Algorithms for inferring cis-regulatory structures and protein interaction networks
RECOMB '04 Proceedings of the eighth annual international conference on Resaerch in computational molecular biology
| Hi-index | 0.00 |
The complexity of the global organization and internalstructures of motifs in higher eukaryotic organisms raisessignificant challenges for motif detection techniques. Toachieve successful de novo motif detection it is necessary tomodel the complex dependencies within and among motifsand incorporate biological prior knowledge. In this paper,we present LOGOS, an integrated LOcal and GlObal motifSequence model for biopolymer sequences, which providesa principled framework for developing, modularizing,extending and computing expressive motif models forcomplex biopolymer sequence analysis. LOGOS consistsof two interacting submodels: HMDM, a local alignmentmodel capturing biological prior knowledge and positionaldependence within the motif local structure; and HMM, aglobal motif distribution model modeling frequencies anddependencies of motif occurrences. Model parameters canbe fit using training motifs within an empirical Bayesianframework. A variational EM algorithm is developed for denovo motif detection. LOGOS improves over existing modelsthat ignore biological priors and dependencies in motifstructures and motif occurrences, and demonstrates superiorperformance on both semi-realistic test data and cis-regulatorysequences from yeast and Drosophila sequenceswith regard to sensitivity, specificity, flexibility and extensibility.