Eight K-means clusters (see additional files 1, 2, 3, 4, 5, 6, 7 and 8: Heat maps of the generated Clusters A to H; additional file 9: GS-9973 purchase combined spread sheet of the clustered genes) was found to be the smallest number resulting in clusters with clearly distinguishable expression characteristics. These expression characteristics become apparent by calculating the arithmetic mean expression profile of all contained genes (Fig. 2). Figure 2 The eight clusters of the transcriptomic buy MK0683 profiling

of S. meliloti 1021 following a shift to acidic pH. The diagrams of clusters A to H show the mean M value (y-axis) obtained by the Sm6kOligo microarray analyses for each time point (x-axis) after pH shift. The standard deviations are represented by the vertical lines crossing each point of the graph. The dotted line divides the cellular response into two parts. The location of the dotted line was chosen according to the observation that most

of the cellular response happened in the first 20 minutes. The cluster analysis generated different groups with discrete expression profiles for up-regulated genes (cluster A to D in Fig. 2) and down-regulated genes (cluster E to H in Fig. 2), not only differing in the intensity of HSP inhibitor drugs their expression level, but also in their time dependent expression behaviour. These time dependent behaviours can be roughly separated into clusters containing genes that were permanently differentially expressed and clusters containing genes which were only transiently differentially expressed (Fig. 3). An inspection of individual expression profiles (data not shown) indicated that for borderline cases the passage between clusters is fluent. The mean expression profiles of the clusters indicated that the main changes in response to low pH occurred approximately within the first 20 minutes (Fig. 2). After this period of time a constant differential expression level or a constantly

changing differential expression level can be observed for Elongation factor 2 kinase most clusters. It is also noticeable that several clusters contain genes organised in operons and groups of genes belonging to related or similar cellular functions. Figure 3 Grouping of S. meliloti 1021 genes following a shift to acidic pH. The eight calculated gene clusters were characterised by their specific transcriptomic response. The figure shows the classification of the genes chosen for clustering by single attributes into the eight clusters calculated by K-means. The tables below give the gene names of the genes distributed to the corresponding clusters. In cluster A, genes exhibiting a strong and permanent induction accumulated. Genes in this cluster remained up-regulated for the whole observation period. It therefore seems that these genes have a special impact for S. meliloti in facing low pH conditions. Clusters B also contains genes that remained permanently up-regulated in response to the pH shift, but not as strong as those in cluster A.