Abstract
Supplementary Table 7 shows sample-level clinico-molecular outcome annotations used to train and test all models, provided in the clinico-molecular_annotations sheet. The Top100_CpGs sheet lists the 100 CpGs most strongly associated with each driver mutation or mutation burden (rows = CpGs; columns = driver mutations or mutation burden), corresponding to the data shown in Supplementary Figure 14g. Variable importance scores for the gene-level methylation predictors used in each predictive model (separate sheet per model), scaled from 0 (least important) to 100 (most important)