Date of Award


Degree Type


Degree Name

Master of Science (MS)


Mathematics and Statistics

Committee Chair(s)

John R. Stevens


John R. Stevens


D. Richard Cutler


Daniel C. Coster


There are many platforms available for simultaneously measuring the relative activity, or expression, levels of all genes in an organism. Genes that have systematically dier- ent expression levels between experimental factor levels are called \dierentially expressed". Because genes are annotated based on their known roles in biological processes (BP), molec- ular functions (MF) and cellular components (CC), gene expression levels can be used to determine relative activity levels of individual BP, MF, or CC between experimental fac- tor levels (this is called gene set testing). Often multiple experimental dierences are of interest simultaneously, which necessitates multivariate gene set testing. Only genes that contribute to a BP, MF, or CC are annotated to that gene set, and it is necessary to have a consensus of genes contributing for a BP, MF, or CC to function properly. Because of this, the direction of the dierential expression (up of down regulated) matters. mvGST, which is platform and design independent, includes tools to summarize and visualize results from multivariate and directional gene set tests. This exibility will be demonstrated on two examples. The rst example comes from a study evaluating the eect of dierent types of fertilization on pig embryos at dierent stages in the embryos' development. The second example comes from a publicly available gene expression data set from a study evaluating the eect of dierent chemotherapy dosages on dierent leukemia blood cell lines. There are plans to make mvGST available publicly through CRAN (Comprehensive R Archive Network) or Bioconductor [1].