HighProbability determines which alternative hypotheses are sufficiently probable: Genomic applications include detection of differential gene expression
Abstract
Many genomic experiments, notably microarray experiments seeking to detect differential gene expression, involve calculating a large number of pvalues. This leads to the multiple testing problem: when the number of null hypotheses is large, the probability of accepting at least one false alternative hypothesis is often much greater than the significance level of the tests, which tends to mislead investigators. Software called HighProbability provides a simple, fast, reliable solution to the multiple testing problem, with applications to many areas of bioinformatics. For example, in a microarray study, HighProbability can determine which genes are probably differentially expressed. Given a set of pvalues not adjusted for multiple testing, HighProbability determines which ones are low enough to imply a high probability of the truth of their alternative hypotheses. The set of pvalues may be determined by conventional hypothesis testing or by random permutations using existing R or SPLUS software. HighProbability is freely available under license through http://www.davidbickel.com . Coded in S, HighProbability currently requires an installation of R or SPLUS, but the algorithm is short enough for fast implementation in nonS languages as well.
 Publication:

arXiv eprints
 Pub Date:
 February 2004
 arXiv:
 arXiv:qbio/0402049
 Bibcode:
 2004q.bio.....2049B
 Keywords:

 Quantitative Biology  Quantitative Methods;
 Quantitative Biology  Molecular Networks;
 6207;
 62P10
 EPrint:
 Submitted for publication