Discovering trends in gene expression data using a hybrid evolutionary algorithm

High–throughput technology has enabled molecular biologists to study genes and gene products of living organisms on a systems level: nowadays, it is possible to measure the activity of thousands of genes in a single experiment. With this type of measurement, one aims at revealing the structure and the dynamics of the underlying genetic regulatory network. In particular, one is interested in identifying groups of genes with shared functions or shared regulatory mechanisms which leads to various challenging optimization problems. Here, we consider the problem of finding multiple, diverse modules of genes that exhibit similar trends regarding one or several gene expression data sets. We present a hybrid evolutionary algorithm for this task that distinguishes itself from previous approaches in three aspects: (i) a set of diverse modules can be found in a single optimization run, (ii) multiple data sets can be considered simultaneously without mixing the corresponding data, and (iii) the trade–off between available runtime and quality of the generated solution can be set by the user.