Figure 4
Figure 4. Methylation or expression level or both of 16 genes can accurately differentiate between ABC and GCB DLBCLs. (A) Sixteen genes represent an overlap between differentially methylated (P < .001) and most differentially expressed (P < .001) genes between ABC and GCB DLBCLs. This set of genes represents genes potentially regulated by methylation with inverse methylation and expression levels (the inverse correlation is not random with P < .005 in Fisher exact test). (B-C) The mean log-ratio levels in ABC vs GCB DLBCLs of methylation and expression values of the 16 signature genes are depicted according to color scale in the heatmaps. The prediction probability graphs (on top of heatmaps) show that 16-gene methylation level can assign ABC and GCB labels with 92% accuracy with the use of Bayesian predictor at probability cutoff of 0.8 (B), whereas gene expression level can assign ABC and GCB labels with 98% accuracy with the use of Bayesian predictor at probability cutoff of 0.8 with an independent set of 203 DLBCL cases from Lenz et al11 (C).

Methylation or expression level or both of 16 genes can accurately differentiate between ABC and GCB DLBCLs. (A) Sixteen genes represent an overlap between differentially methylated (P < .001) and most differentially expressed (P < .001) genes between ABC and GCB DLBCLs. This set of genes represents genes potentially regulated by methylation with inverse methylation and expression levels (the inverse correlation is not random with P < .005 in Fisher exact test). (B-C) The mean log-ratio levels in ABC vs GCB DLBCLs of methylation and expression values of the 16 signature genes are depicted according to color scale in the heatmaps. The prediction probability graphs (on top of heatmaps) show that 16-gene methylation level can assign ABC and GCB labels with 92% accuracy with the use of Bayesian predictor at probability cutoff of 0.8 (B), whereas gene expression level can assign ABC and GCB labels with 98% accuracy with the use of Bayesian predictor at probability cutoff of 0.8 with an independent set of 203 DLBCL cases from Lenz et al11  (C).

Close Modal

or Create an Account

Close Modal
Close Modal