This website provides supplemental information for the following papers:
R.O. Dror, J.G. Murnick, N.J. Rinaldi, V.D. Marinescu, R.M. Rifkin, and R.A. Young. A Bayesian approach to transcript estimation from gene array data: the BEAM technique. Proceedings of the Sixth Annual International Conference on Research in Computational Molecular Biology , Washington, DC, April 2002.
R.O. Dror, J.G. Murnick, N.J. Rinaldi, V.D. Marinescu, R.M. Rifkin, and R.A. Young. Bayesian Estimation of Transcript Levels Using a General Model of Array Measurement Noise. To appear in Journal of Computational Biology, 2003.
The latter paper presents a more complete description of the method and uses an improved noise model. You may also find the supplement on Efficient Computation of BEAM Estimates (PDF format) useful.
If you wish to apply the BEAM method using the noise model (and parameter values) presented in the latter paper, you can use our precomputed lookup tables. MATLAB code and data for computing estimates are available as a zipped archive file containing three MATLAB functions (.m files), and nine data files stored in ASCII format. These are all text files, and the software will run in MATLAB on any platform (Windows, Unix, Linux, Mac, etc.). The files can be browsed and downloaded individually.
Additionally, we have implemented a web-based version of BEAM --- just type observed expression values into the form, and see the BEAM estimates. The BEAM software currently includes the following three functions:
beam1: estimated expression level from a single observation. [Browse Code][Try It] beam2: estimated expression level from two measurements under the same condition. [Browse Code][Try It] beamr: estimated ratio of expression levels of two measurements under different conditions. [Browse Code][Try It]
All three functions load their own data files (lookup tables) as they need them, and work with vectors of estimates. For more information, see the help for each individual function.
The data files available on this Web page assume a noise model derived from Affymetrix Ye6100 yeast gene chips. Although we show in the Journal of Computational Biology paper that this model provides a reasonable fit to data from the U95A human chip, you may wish to derive a noise model specific to the array you are using, especially if you are using cDNA arrays. Methods for deriving noise and prior models from experimental data are described in the papers.