Fig. 10

Results of manual cluster refinement and comparison with additional data types and clustering results. a The original clustering is shown in the leftmost column. b The records in the clusters are sorted by within-cluster distances, c the bottom sections of the clusters contain the expression profiles with poor cluster fit. d The second column from the left shows the manually refined clustering with 5 clusters. e Here, poorly fitting records from two groups of the original clustering were split off and subsequently merged. f The Kaplan-Meier plots show patient survival for each of the five clusters in the second column. g The three rightmost columns are k-Means clustering results computed with the software. Notice that the manually derived cluster e consistently appears in the k-means clustering results