Microarray Toolbox

PowerExpress

    PowerExpress is a tool for selecting genes from microarrays. The main features of PowerExpress include:

(1) PowerExpress can be used to search for genes that show complex expression patterns of interest (e.g. genes that show specific temporal or spatial patterns). In PowerExpress, one can specify a complex criteria for conducting multiple sample comparisons such as "(mutant 1 < wild type < mutant 2) and [(mutant 3 > wild type) or (mutant 4 <mutant 5)]", and the program will then automatically rank the genes according to the specified criteria.

(2) PowerExpress is especially useful for dealing with small number of replicates. The software adopts an hierarchical empirical Bayes estimator for variance estimation. It pools information from all genes across the array and can significantly increase the sensitivity in gene selection when there are only a few replicates available.

(3) Multiple probes or probesets of a gene can be combined to make inference about the gene.

    PowerExpress consists of two parts: powexpress and powexpressloc. Powexpress is used to select probes/probesets that show specific expression patterns. Powexpressloc is used to select genes that show specific expression patterns by combining probe/probset level test-statistics computed by powexpress.
.
    The algorithm used by PowerExpress is described in:

        Ji, H. and Wong, W.H. (2005) Increasing Power of Gene Selection from Microarrays: A Hierarchial Empirical Bayes Approach. (Submitted).

    The executable version can be downloaded here:

        powexpress:

                (a) Readme

                (b) Executables: Windows, Mac OS X

                (c) Sample Input Files: data file, annotation file, parameter file

        powexpressloc:

                (a) Readme

                (b) Executables: Windows, Mac OS X

                (c) Sample Input Files: parameter file, sample *.ori file

        Academic users who want to run PowerExpress on other platforms can email ji@fas.harvard.edu to get source code in ANSI C.

    Supporting data and supplementary materials for the paper above can be downloaded here:

    (a)  Embryonic Stem Cell Data: EB8, OCT4

    (b)  Shh Data

            For this data, gene identities are masked for temporary and will be released upon the publication of corresponding biological studies. The array labeled by "*" is contaminated by scanning noise. The coded Locuslink ID can be used to group probesets into genes, but they are not the real Locuslink ID.

    (c)  ChIP-Chip Tiling Array Data

    Future:

    Currently, we don't provide a Graphic User Interface. However, as part of CisGenome project, a GUI will be eventually developed for PowerExpress as well as other tools provided in this website.