Figure 4

Accurate clustering of the trisomy 8, disomy 8, and reference cultures by PCA and supervised HCA of genome-wide methylation patterns and significant hypomethylation of gene-poor regions on chromosome 8 only in the trisomy 8 cultures. (A) The PCA identified 781 BACs that displayed significant differences between the trisomy 8, disomy 8, and reference cultures (left). Supervised HCA of these BACs accurately clustered the subgroups, as illustrated in the heat map (right). (B) A similar pattern was observed when analyzing only chr8 data, with the PCA identifying 81 BACs (left) and HCA accurately clustering the subgroups (right). (C) Mean log2 ratio of the three culture subgroups plotted against the number of genes on the BACs for chr8 revealed a significant hypomethylation of gene-poor regions in trisomy 8 cultures. The black curve shows the proportions that the BAC clones with different number of genes constitute of all BAC clones on chr 8; for example, the BACs with no genes comprise 45% of all the BAC clones on this chromosome. (D) An independent and weighted ANOVA and a Tukey HSD test verified the global hypomethylation of gene-poor regions on chr8 in trisomy 8 cultures but not in the disomy 8 or reference cultures. (E) Mean log2 ratio of the three culture subgroups plotted against the number of genes on the BACs for chromosome 7 (used as an example of another autosome) revealed no significant differences in genome-wide methylation. (F) Mean log2 ratio of female and male cultures plotted against the number of genes on the BACs for chromosome X revealed significant hypomethylation of gene-poor regions only in the female culture. ANOVA, analysis of variance; BAC, bacterial artificial chromosome; chr8, chromosome 8; HCA, hierarchial cluster analysis; HSD, honestly significant difference; PCS, principal component analysis.