skip to primary navigationskip to content

BioHMM: A heterogeneous hidden Markov model for segmenting array CGH data

John Marioni, Natalie Thorne and Simon Tavaré


We have developed a new function called BioHMM which uses a heterogeneous hidden Markov model to segment array comparative genomic hybridisation data. Such a model makes use of covariate information such as the distance between clones or clone quality in the segmentation. By using such information, we provide the user with the opportunity to take account of information which is not used by any of the current segmentation schemes.

Supplementary material for the paper on BioHMM can be downloaded from here. This document describes the extension of BioHMM to more than two states and contains a discussion about how multiple covariates might be incorporated in the model. Additionally, we provide an example of the format in which data must be read into BioHMM as well as several examples (and discussion) of the output obtained upon applying BioHMM. This is a modified version of the supplementary material that was reviewed. It has been altered to take account of changes to the snapCGH library. The Original supplementary material can be downloaded from here.

BioHMM is available as part of the snapCGH library, which can be downloaded from the BioConductor developer website here.

Practicals describing the use of snapCGH for some small datasets can be found at Practical I and the data required to run the practical can be found at Practical I data as a zip file.

For more details of BioHMM please contact Dr. John Marioni