Segway 2.0: Gaussian mixture models and minibatch training.

Rachel C W Chan,Eric G Roberts,Michael M Hoffman,Jeffrey A Bilmes,William Stafford Noble,Maxwell W Libbrecht,Inanc Birol

doi:10.1093/bioinformatics/btx603

Abstract

SummarySegway performs semi-automated genome annotation, discovering joint patterns across multiple genomic signal datasets. We discuss a major new version of Segway and highlight its ability to model data with substantially greater accuracy. Major enhancements in Segway 2.0 include the ability to model data with a mixture of Gaussians, enabling capture of arbitrarily complex signal distributions, and minibatch training, leading to better learned parameters.Availability and implementationSegway and its source code are freely available for download at http://segway.hoffmanlab.org. We have made available scripts (https://doi.org/10.5281/zenodo.802939) and datasets (https://doi.org/10.5281/zenodo.802906) for this paper’s analysis.Supplementary information Supplementary data are available at Bioinformatics online.

Highlights

Segway identifies recurring combinatorial patterns in multiple genomewide signal datasets such as ChIP-seq or DNase-seq data (Hoffman et al, 2012)
Availability and implementation: Segway and its source code are freely available for download at http://segway.hoffmanlab.org
Contact: michael.hoffman@utoronto.ca Supplementary information: Supplementary data are available at Bioinformatics online

Summary

Introduction

Segway identifies recurring combinatorial patterns in multiple genomewide signal datasets such as ChIP-seq or DNase-seq data (Hoffman et al, 2012). The labels might represent genomic features such as ‘enhancer’ or ‘facultative heterochromatin’. We expect the new standalone mode to interest the most users. This mode removes the requirement for a cluster system such as Sun Grid Engine, allowing one to run Segway on any Linux host. Of interest are new features which improve Segway’s ability to learn more complex patterns with less configuration. We describe these features and demonstrate the improvement they provide

Minibatch training

Gaussian mixture models

Comparison with other methods