Fig. 1

Cohort selection. Unsupervised clustering of gene expressions was used to inspect how LGGs and GBMs clustered. a Expressions on the full cohort were clustered according to the 500 genes most varied across samples. b Expression of Cheng et al.’s 64-gene was averaged to represent a “metagene”; The (69) samples in the lowest metagene-expression quartile became the “Low MS” cohort, whereas the (69) samples in the highest metagene-expression quartiles became the “High MS” cohort