Agreement with standard gene classes. Perplexities were computed for all methods for two datasets: Osmotic shock response, left) and DNA damage, right). We used three standard gene cluster sets: (A) All genes, (B) Common genes, (C) Total genes. The 2SE error bars are over 20 runs. Lower perplexity score is better. Mat is Matisse.
Parkkinen and Kaski BMC Systems Biology 2010 4:4 doi:10.1186/1752-0509-4-4