Assessment of the gene expression clustering results using the Computational Negative Control (CNC) approach. (A) For each specified number of clusters, the cluster quality metric, silhouette coefficient (SC), is evaluated and compared to that from the randomly permuted data. (B) Difference in SC from (A) multiplied by number of clusters shows a marked decrease at more than six clusters, indicating that SC is no longer distinct from the randomized data.
Juskeviciute et al. BMC Genomics 2008 9:527 doi:10.1186/1471-2164-9-527