B0584
Title: AlertGS: Calculating alerts for gene sets based on individual dose-response modelling
Authors: Franziska Kappenberg - TU Dortmund University (Germany) [presenting]
Joerg Rahnenfuehrer - TU Dortmund University (Germany)
Abstract: A typical pipeline for dose-response experiments with gene expression as a response variable is to perform differential expression analysis for each dose against a negative control individually. Based on the significant genes found by this approach, overrepresentation analysis of for instance Gene Ontology groups is often performed. However, this does not allow interpolation between the actually measured dose values, because the dose is considered only as a qualitative factor. Recent research suggests that calculating alerts, i.e. the dose value where some pre-specified effect of interest is attained or significantly exceeded, is reasonable, especially for gene expression data. A new approach is suggested to directly estimate an alert for an entire gene set based on the individual model-based alerts for the genes contained in that set. The method uses a Kolmogorov-Smirnov test for the ordered alerts of all genes. Significance statements are made via permutations of the alerts. This method is evaluated for a specific sample dataset, and results from the new approach are compared to results from established approaches for calculating enrichment scores based on the differential expression analysis. In addition, the performance of the method is assessed in some simple simulation scenarios.