Monte Carlo Null Models for Genomic Data

• Activity Number: 127
• Type: Contributed
• Date/Time: Monday, August 5, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Egil Ferkingstad*+ and Lars Holden and Geir Kjetil Sandve
• Companies: Norwegian Computing Center and Norwegian Computing Center and University of Oslo, Department of Informatics
• Keywords: monte carlo methods ; genomics ; hypothesis testing

Abstract

As increasingly complex hypothesis-testing scenarios are considered in many scientific fields, analytic derivation of null distributions is often out of reach. To the rescue comes Monte Carlo testing: as long as you can sample test statistics under the null hypothesis, the p-value is just the proportion of sampled test statistics that exceed the observed test statistic. Sampling test statistics is often simple once you have defined a Monte Carlo null model for your data, and specifying some form of randomization procedure may also be relatively straightforward. However, there may be several possible choices of null model for the problem, and no clear-cut criteria for choosing among them. Unrealistically low p-values may thus occur due to the inadequacy of the chosen null model. We show that, in many cases of interest, we may order null models by increasing preservation of data characteristics, and this implies a corresponding ordering of statistical significance. We illustrate our findings on genomic data using the Genomic HyperBrowser, a generic web-based system providing statistical methodology and computing power to handle a variety of biological inquiries on genomic datasets.