Supplementary Figure 8: Reconstruction of the phyloepigenetic trees with mitigation for the effects of immune cell content.

We recalculated each phyloepigenetic tree using one of two methods to mitigate the effects of immune cells within the samples. For the first method, immune cell adjusted, we estimated the fraction of leukocytes in each sample using profiles of immune-specific methylation probes, as described previously (Nat. Biotechnol. 30, 413–421, 2012; Nat. Genet. 45, 1134–1140, 2013). We then used the methylation profile of pure leukocytes, along with the estimated leukocyte/cancer cell mixture, to infer the cancer cell methylation value for every probe (probe numbers are not identical because of different probes being below the detection limit in different experiments). For the second method, dichotomized, we used only probes that had a fully unmethylated state in pure leukocytes and dichotomized/binarized each probe on the basis of the minimum methylation level in tumor samples (Online Methods). For each of the immune cell adjusted and dichotomized trees, the RF distance shows similarity to the reference version.