GSAR

"GSAR" (Gene Set Analysis in R) introduces an innovative approach to analyzing differentially expressed gene sets in gene expression data, focusing on the use of minimum-spanning tree (MST)-based non-parametric multivariate tests. While a wide range of tests are available for gene set analysis, including aggregation tests and multivariate tests that consider intergene correlations, many of these tests lack specificity in identifying the exact nature of the deviation from the null hypothesis upon rejection.

The core contribution of GSAR is its comparison of the power and Type I error rates of MST-based tests with other commonly used tests in pathway analyses. Through simulation studies, GSAR demonstrates that MST-based tests have strength comparable to conventional tests in many scenarios and offer superior performance in specific parameter spaces that are biologically relevant. These tests are particularly effective in distinguishing between shift and scale alternatives, a critical aspect of analyzing gene expression data.

GSAR proposes a two-step practical analysis strategy to enhance the interpretability of experimental data. The first step involves applying a powerful multivariate test to identify gene sets where the null hypothesis is rejected. The second step uses MST-based tests on these identified pathways to pinpoint those supporting specific alternative hypotheses. This approach not only refines the analysis process but also clarifies the underlying biological phenomena.