Affiliations: Department of Computer Science, Ben-Gurion University,
Be'er-Sheva, Israel | Department of Virology and Molecular Development,
Soroka University Medical Center, Faculty of Health Sciences, Ben-Gurion
University, Be'er-Sheva, Israel
Note: [] Corresponding author: Avraham A. Melkman, Department of Computer
Science, Ben-Gurion University, Be'er-Sheva, 84105, Israel. E-mail:
[email protected]
Abstract: The low reproducibility of differential expression of individual
genes in microarray experiments has led to the suggestion that experiments be
analyzed in terms of gene characteristics, such as GO categories or pathways,
in order to enhance the robustness of the results. An implicit assumption of
this approach is that the different experiments in effect randomly sample the
genes participating in an active process. We argue that by the same rationale
it is possible to perform this higher-level analysis on the aggregation of
genes that are differentially-expressed in different expression-based studies,
even if the experiments used different platforms. The aggregation increases the
reliability of the results, it has the potential for uncovering signals that
are liable to escape detection in the individual experiments, and it enables a
more thorough mining of the ever more plentiful microarray data. We present here a proof-of-concept study of these ideas, using ten
studies describing the changes in expression profiles of human host genes in
response to infection by Retroviridae or Herpesviridae viral families. We
supply a tool (accessible at www.cs.bgu.ac.il/∼waytogo) which enables the
user to learn about genes and processes of interest in this study.
Keywords: Data integration, gene expression, microarray data, Gene-Ontology
