Figure 4.

Genes with strong downregulation cluster within and flanking domains with high L1 density on chromosome 17. (a) Genome Environment Browser (GEB) histogram display showing distribution of chromosome 17 genes with ≥ 30% downregulation compared with all downregulated genes. Gene density (green) and full-length L1 (FL-L1) density (light blue) are also shown. Boxes highlight >30% genes enriched in high L1 domains flanking a large gene-rich region. (b) GEB detail view of Mb 18-58. High L1 density (HL1) domains are visualised as regions with high density of vertical lines (L1 5' untranslated region (UTR), open reading frame (ORF)1, ORF2 and 3' UTR) and are highlighted for illustrative purposes with horizontal red lines. Genes are represented as green boxes. Those showing >30% and <30%downregulation are blue and black, respectively. Downregulated classes are also highlighted by black (>30%) and grey (<30%) bars above and below the GEB display. (c) Classification tree comparing 20% most downregulated genes on chromosome 17 with all other genes. The first split divides the chromosome into two, proximal and distal of 56 Mb, based on the distance between the gene and a unique L1VL4 element at 80.4 Mb (×60). On the right side of the tree, representing the proximal 56 Mb, the model suggests genes located in HL1 domains larger than 0.54 Mb (×31) had a much higher probability of being silenced. (d) Regression tree for chromosome 17 highlights equivalent L1 domain related features. At the terminal nodes, the lower the expression values, the stronger the gene silencing effect. The tree defines that to remain active the gene should be (1) within the first 56 Mb of the chromosome (node 1 split, ×82), (2) associated with a relatively small HL1, or no HL1 at all (node 2 split, ×31), and (3) associated with a low L1 density (LL1) domain at least approximately 5.3 Mb large (node 3 split, ×30). Two large LL1 domains on chromosome 17 fit this description. Y-axis scale on box and whisker plots shows expression level of genes relative to untreated control in log2 scale.