Fig. 1

Deregulation of chromatin regulators in OSCC. A Volcano plots showing the 34 DEGs between normal (n = 29) and OSCC (n = 281) samples in the TCGA database and clarifying the 19 chromatin regulators were closely related to OSCC tumorigenesis by Lasso logistic regression and Boruta machine learning algorithm. B Differential expression profiles of 14 chromatin regulators in normal (n = 108) and OSCC (n = 402) samples from the meta-GEO dataset (SP140 was not present on the microarray chip). C Univariate Cox regression analyses of the 14 chromatin regulators in the meta-GEO dataset. Hazard Ratio and P-values were displayed. D, E Random survival forest analysis, where the error rate was reduced to a stable value as the number of trees increases, and genes were ranked according to importance. F Multivariate Cox regression analyses of the 7 chromatin regulators in the meta-GEO dataset. Hazard Ratio and P-values were displayed. G–J Boxplot indicating SMYD3 expression in different tissue samples and cell lines from the TCGA database, GSE37991 dataset, GSE30784 dataset and GSE146483 dataset. Numbers in parentheses indicate the sample size. K–N Up-regulation of SMYD3 had a significantly shorter OS time and PFS time in the meta-GEO dataset, ICGC database and TCGA database. Numbers in parentheses indicate the sample size. P-values were obtained from the log-rank tests. Ns, not significant, *P < 0.05, **P ≤ 0.01, and ***P ≤ 0.001