Modeling gene expression data via positive Boolean functions

In this work we propose an artificial model for the generation of biologically plausible gene expression data to be used in the evaluation of the performance of gene selection and clustering methods. The model allows to fix in advance the set of relevant genes and the functional classes involved in the problem; the input-output relationship is constructed by synthesizing a positive Boolean function. Despite its simplicity, it is sufficiently rich to take account of the specific peculiarities of gene expression data, including biological variability. A Java code had been developed to allow the user choose the model parameters according to the characteristics of the experiment he want to simulate. This permits to insert the artificial model into a distributed system for microarray analysis, in particular one based on a Grid infrastructure.