A Biclustering Analysis Tool implementing the BiHEA Algorithm

GALLO, CRISTIAN ANDRÉS; DUSSAUT, JULIETA SOL; CARBALLIDO, JESSICA ANDREA; PONZONI, IGNACIO

Bernal

Congreso; 1er Congreso Argentino de Bioinformática y Biología Computacional; 2010

Asociación Argentina de Bioinformática y Biología Computacional

Microarray technology has become a central tool in biological research, and the identification of gene groups with similar expression patterns represents a key step in the analysis of gene expression data. Traditional clustering algorithms partition an expression matrix into submatrices that extend over the whole set of conditions. However, in most cases, the assumption that all genes behave similarly in all conditions may be too restrictive. To account for this, biclustering approaches carry out the grouping in both dimensions simultaneously: genes and conditions. This allows to find subgroups of genes that show the same response under a subset of conditions, e.g. if a cellular process is only active under these conditions. Furthermore, if a gene participates in multiple pathways that are differentially regulated, one would expect this gene to be included in more than one cluster; this cannot be achieved by traditional clustering. The BiHEA Algorithm [1] is a Hybrid Evolutionary Algorithm for Biclustering of gene expression data focused on biclusters with coherent values that follows an additive model. In this work, a user friendly software that implements the BiHEA Algorithm and several tools for biclustering analysis is presented. The main features of the software can be summarized as follows: Data handling. All data, being the entire expression matrix loaded from an external file or sets of biclusters, are organized in a list structure that is depicted in the left panel of the graphical user interface (cf. Figure 1). All the work performed with the software can be saved as a project form and restored later.Data pre-processing. The input data file can be any CSV file including annotations of genes and conditions. The loaded gene expression data can then be transformed by means of translation, escalation or standardization. Visualization. The expression matrix can be displayed as a heatmap or as a numerical matrix. Annotations of the conditions run along the top whereas the annotations of the genes are listed on the left hand side. Biclusters can be visualized in three different ways: in the form of a heatmap, a numerical sub-matrix, or as a collection of gene expression profiles (See Figure 1). Post-processing. For further investigations, the software offers the possibility of a gene pair analysis. In particular, for each pair of genes it is calculated how often these genes occur together in the same bicluster. This number of co-occurrence indicates which genes may be functionally related. Additionally, the degree of coverage of the resulting biclusters with respect to the gene expression matrix can be visualized as a grey scale matrix.All the figures, graphs, and results can be exported for further usages on presentations, papers, etc. The BiHEA software is freely available at http://lidecc.cs.uns.edu.ar/BiHEA and runs on all operating systems with a Java Virtual Machine.