Fig. 4

a Hierarchical clustering of 30 R-specific probe sets. The expression datasets were gene-wise normalized and, the clusters for 30 probe sets and four datasets were made using Pearson correlation coefficients. Heatmap showed one vertical cluster and two single nodes for four datasets. Three horizontal clusters were made for 30 probe sets. b Relevance network of 30 R-specific probe sets. Six relevance sub-networks were generated using 30 R-specific probe sets. Sub-networks showed that the expression of one probe set affected the expression of the other probe set either positively (shown in red line) or negatively (shown in blue line). The expression profiles of these probesets among resistant datasets (US-897, RL 5, 17 and 27 wai) are shown in rectangle boxes