Software - Proprietary:
The Bioinformatics Group releases software under two separate licensing
methods. This page links to proprietary software releases.
For open source software, click here.
Bayesian Decomposition is a pattern recognition algorithm which uses Bayesian statistics and Markov chain Monte Carlo sampling to determine the physically significant basis vectors for a data set. Mathematically a matrix factorization is performed, creating a matrix defining the basis vectors (patterns) and the distributions of the patterns within the data. The Bayesian statistics allow for incorporation of prior knowledge through the encoding of correlations between data points. The software is presently available in a form for spectroscopy, a form for studying exponential distributions, and a form which encodes no correlations between data points.
Application of Bayesian Decomposition for analysing microarray data.
T. D. Moloshok, R. R. Klevecz, J. D. Grant, F. J. Manion, W. F. Speier, IV, and M. F. Ochs.
Bioinformatics 18: 566-575, 2002.
This extension to Bayesian Decomposition allows knowledge of class membership to be included.
Bayesian Decomposition with Supervised Learning
Bayesian Decomposition classification of the project normal data set
T. D. Moloshok, D. J. Datta, A. V. Kossenkov, M. F. Ochs
in K. Johnson and S. Lin, Editors, Methods of Microarray Data Analysis III, 211 Ð 232, Kluwer Academic, Boston, 2003.
This extension allows the user to define expected coregulated sets of genes, such as linked by a transcription factor (TF), which BD will use as prior information.
BD-TF Bayesian Decomposition with Prior Knowledge of Coregulation
Determining transcription factor activity from microarray data using Bayesian Markov Chain Monte Carlo sampling
A. V. Kossenkov, A. J. Peterson and M. F. Ochs
Medinfo 12 2007