Skip to main content
. 2015 Nov 6;10(11):e0141695. doi: 10.1371/journal.pone.0141695

Fig 2. Discovery of OSCC subtypes based on miRNA expression among 305 OSCCs in the TCGA dataset.

Fig 2

A) Nonnegatiive matrix factorization consensus clustering showed support for two and four OSCC subtypes based on the maximal cophenetic value at K = 2 and 4. B) Consensus matrices for the samples were calculated at k = 2, 3, 4, 5, 6, 7, 8. A comparison of the visualized consensus matrices allowed a visual assessment of the best rank. Clear block patterns along the diagonal of the consensus matrices indicated robustness of the clustering. C) Analysis of the ever smokers and never smokers in both miRNA based subclass revealed a lack of a difference in the proportions of the two risk groups based on smoking experience, p < 0.059 by Fisher Exact Test D) Analysis of the ever smokers and never smokers in each of 4 miRNA based subclasses revealed no significant difference in proportions of ever and never smokers in any group.