Page 75 - Molecular features of low-grade developmental brain tumours

DISTINCT DNA METHYLATION PATTERNS IN SEGA IN TSC
robust indicated by delta area plot (b), consensus Cumulative Distribution Function (CDF) plot (c) and consensus matrix for k=3 (d). e. Silhouette plot showing the average silhouette with for each k, indicating that k=3 is most robust. f. Barchart of the silhouette with for k=2 showing which samples cluster together. g. Barchart of the silhouette with for k=3 showing which samples cluster together.
128 genes and performed correlations between the normalized count matrix of genes that were expressed (106/128 genes) and the β-values of the corresponding CpGs. We identified 11 genes (MFAP5, SLC15A2, GPT2, CLCC1, CYP4F12, KLHL33, C12orf29, SELENBP1, MTERFD1, ICAM2, CHD7) that inversely correlated with their corresponding CpG (Table 3). Based on the RNA expression of the expressed genes no clear clustering was found between SEGA1 and SEGA2, although the number of cases in each group was relatively small (Figure 5f).
Next, we investigated if the expression of CD3, HLA-DP/DQ/DR, GFAP, MAP2 or pS6 could explain the subgroups found in the methylation data. We found a higher positive area of CD3 in SEGA2 compared to SEGA1 (p= 0.0068; Figure 6a boxplot). No difference was found between SEGA1 and SEGA2 for the other markers (Figure 6b-6e; boxplots) or between the subgroups of SEGA2, SEGA2a and SEGA2b (Figure 6a-e; boxplots). Using ROC analysis we identified CD3 positive area as the best predictor for dividing SEGA1 from SEGA2 (Figure 6a; AUC=0.825) compared to HLA-DP/DQ/DR (Figure 6b; AUC=0.561), GFAP (Figure 6c; AUC=0.450), MAP2 (Figure 6d; AUC=0.481) and pS6 (Figure 6e; AUC=0.539). Furthermore, using Random forest we found that none of the clinical data (TSC1/TSC2 mutation status, gender, localization of the SEGA, age at seizure onset, seizure frequency, drug management at time of surgery, size of the tumour, tumour recurrence/regrowth and presence of other TSC related malformations) could properly separate between SEGA1 and SEGA2 (Figure 6f) or between SEGA1, SEGA2a and SEGA2b (Figure 6g). The tumour size contributed most to separating the two groups but showed no significant difference between SEGA1 and SEGA2 (Figure 6h).
Table 2. Top 50 GO terms enriched in SEGA compared to control tissue.
GO term
GO
p. value
p. adjust
q. value
genes (n)
Hypo
Hyper
hypo & hyper
leukocyte cell-cell adhesion
leukocyte aggregation
T cell activation
regulation of cell activation
leukocyte migration
regulation of mononuclear cell proliferation
GO: 4,05218 0007159 E-21
GO: 4,59205 0070486 E-21
GO: 1,30039 0042110 E-20
GO: 3,05025 0050865 E-19
GO: 1,441 0050900 E-18
GO: 6,18246 0032944 E-18
1,28302 9,57805 163 19 135 9 E-17 E-18
1,28302 9,57805 155 18 129 8 E-17 E-18
1,81665 1,35617 152 17 127 8 E-17 E-17
2,8408 2,12073 160 17 135 8 E-16 E-16
1,15033 8,58752 129 21 101 7 E-15 E-16
3,74405 2,79502 E-15 E-15
81 9 71 1
73
3