In both species, two main expression clusters could be identified that corresponded to up- and down-regulated genes at C2 stage, respectively.
In addition to the 2 main expression clusters characterized by a marked up- (cluster 1) or down-regulation (cluster 2) in C2 follicles, a smaller cluster characterized by an up-regulation in C1 follicles and a subsequent down-regulation in C2 follicles was found in both species (cluster 3, Figure 1A).
The average cutoff value for the main clusters is 0.9; this value is quite high and indicates fuzzy boundaries between many expression clusters.
To analyse the expression profiles of other genes affecting grain quality, not belonging to the main co-expression clusters, we performed GO enrichment analyses of all the DEGs identified in the pair-wise comparisons.
When Verdejo DEG were clustered, only three main expression profile groups were detected because profiles of transcripts up-regulated around noontime could not be distinguished from those up-regulated during the daytime contrasting with the Tempranillo experiment.
Successive clustering with increasing cluster number reflected this finding showing first the main expression patterns and subsequently the minor patterns.
Structuring this data in clusters provided a clear illustration of the main expression profiles of many novel genes, such as genes from the functional classes 'chemotaxis' and 'phage, transposon, or plasmid' which are found in clusters 1 and 5 respectively.
This generated two main clusters (Figure 3A) where patients in the high expression cluster exhibited more rapid disease progression and poorer survival than patients in the cluster with lower expression (Log-rank P values, 0.025 and 0.028) (Figures 3B and 3C).
CXCL9, CXCL10, CXCL11 and CCL5 were co-expressed and, when used for hierarchical clustering analysis based on mean-centered expression values, segregated melanoma metastases into three main groups: cluster 1 included samples with low CXCR3 and CCR5 ligand gene expression; cluster 2 and cluster 3 with intermediate and high expression, respectively.
Clustering of the expression profiles of these 564 genes using a combination of self-organizing maps (SOM) and K-means clustering revealed the presence of a total of 12 co-expression clusters: six main clusters, of which four are divided into sub-clusters.
To get a broader perspective on the function of differentially expressed genes, we analyzed for over-representation of gene ontology (GO) categories in each of the three main gene expression pattern clusters identified in figure 4A.
