Figure 2

Hidden patterns revealed by meta-clusters across muscle-related transcriptome data sets. A- The right-hand side presents a systematic meta-analysis of all the "good" clusters from the MADMuscle database. The results are represented by the diagonally symmetric interaction matrix for 1,247 clusters of co-expressed genes based on statistical comparison (p-values). Red indicates significant similarity (strong overlap), green significant dissimilarity (poor overlap) and grey no overlap between compared clusters. The matrix was ordered using hierarchical clustering (lines and columns) to highlight groups of similar clusters. The resulting visualization uncovers 6 main components of highly similar and interconnected clusters, called meta-clusters (M1 to M6), involved in similar biological processes. B- On the left-hand side, a three-dimensional representation of the functional annotation of the meta-clusters is given. The functional annotation is visualized as a terrain map, in which highly correlated Gene Ontology (GO) terms are placed in proximity in the x-y plane and the probability density of enrichment in a region is shown by the altitude in the z direction. Each line represents a different GO term with cold colors for enrichment values <1 and hot colors for values >1 (see color scale). The rows (7554 GO terms) were ordered by hierarchical clustering, placing terms with similar annotation patterns together. The columns (clusters) were ordered in the same order as in the clustered diagonally symmetric interaction matrix (right-hand side of the figure). Specific significant GO annotation is associated to each Meta-cluster. For instance, the meta-cluster M4 supports genes encoding muscle and mitochondrion proteins.