Extended Data Fig. 5: Associations of MPs with clinical features in TCGA data.

(a) Heatmap showing the hazard ratio of scores for MPs (rows) with respect to overall survival in each cancer type (columns). Purple and green represent association with worse and better survival, respectively. For relevant carcinoma types, the columns of results from data deconvolved by CIBERSORTx are presented with the suffix “(d)”, adjacent to the corresponding column for non-deconvolved data. Significant associations are labelled with ‘*’ for p < 0.05, ‘**’ for p < 0.01 and ‘***’ for p < 0.001. Significance levels were computed via Cox regression and likelihood ratio test, and adjusted by Benjamini-Hochberg correction. Exact p values are shown in Table S8. (b-c) Heatmaps as in a showing the difference in average scores between tumors stratified by lymph node metastasis and therapy resistance, respectively. Significance levels were computed by two-sided t-test and adjusted by Benjamini-Hochberg correction. Exact p values are shown in Table S8. (d) Scatterplot showing consistency of associations of MPs with clinical features in TCGA data. Each point represents an MP, its X-value being the sum of the signs (1 for positive effect, −1 for negative effect) of all significant effects (adjusted p value < 0.05, with adjusted p values as in a-c) for that MP across all cancer types and clinical features, and its Y-value being the sum of the signs of all effects for that MP. Selected points are labelled with their corresponding MPs.