Supplementary Figures S1-S23

Supplementary Figure S1. Mutations data are well-powered to detect sex differences. Power analysis for proportion test. Axes show group sample sizes (male vs female groups). Heatmap colouring shows statistical power for each combination of group sample sizes to detect varying effect sizes. Tumour type-specific sample sizes are indicated.

Supplementary Figure S2. Univariate associations between mutation burden and sex. Pan-cancer sex-biases in (a) somatic mutation prevalence and (b) percent genome altered. Similarly, tumour type-specific sex-biases in (c) somatic mutation prevalence and (b) percent genome altered. Each point represents a sample, where blue points are male-derived tumours and pink points are female-derived tumours. Red lines show the mean mutation burden for each sex, and the FDR-adjusted t-test p-values are labeled. For pan-cancer plots, multivariate p-value is below univariate p-value. All data transformed using Box-Cox power transformations for visualization and statistics.

Supplementary Figure S3. Microsatellite instability and mutation load. (A) Comparison of microsatellite stable (MSI) and instable (MSI-L and MSI-H) vs. sex. (B) Box-Cox transformed mutation load after adjustment for MSI compared between the sexes. Multivariate q-values shown.

Supplementary Figure S4. Sex-biased CNAs in kidney papillary cell cancer. Each plot shows, from top to bottom: the q-value showing significance of sex from multivariate modeling on bins, proportion of samples with aberration, difference in proportion between male and female groups for amplifications; the same repeated for deletions and the CNA profile heatmap. The columns represent ordered by . The rows of the heatmap represent samples with the covariate bars indicating clinical variable included this tumour type-specific model. Supplementary Figure S5. Sex-biased CNAs in head and neck squamous cell cancer. Each plot shows, from top to bottom: the q-value showing significance of sex from multivariate modeling on gene bins, proportion of samples with aberration, difference in proportion between male and female groups for amplifications; the same repeated for deletions and the CNA profile heatmap. The columns represent genes ordered by chromosome. The rows of the heatmap represent samples with the covariate bars indicating clinical variable included this tumour type-specific model.

Supplementary Figure S6. Sex-biased CNAs in stomach & esophageal carcinoma. Each plot shows, from top to bottom: the –log10 q-value showing significance of sex from multivariate modeling, proportion of samples with aberration, difference in proportion between male and female groups for amplifications; the same repeated for deletions and the CNA profile heatmap. The columns represent genes ordered by chromosome. The rows of the heatmap represent samples with the covariate bars indicating clinical variable included this tumour type-specific model. Supplementary Figure S7. Sex-biased CNAs in liver hepatocellular carcinoma. Each plot shows, from top to bottom: the q-value showing significance of sex from multivariate modeling on gene bins, proportion of samples with aberration, difference in proportion between male and female groups for amplifications; the same repeated for deletions and the CNA profile heatmap. The columns represent genes ordered by chromosome. The rows of the heatmap represent samples with the covariate bars indicating clinical variable included this tumour type-specific model.

Supplementary Figure S8. Sex-biased CNAs in bladder urothelial carcinoma. Each plot shows, from top to bottom: the q-value showing significance of sex from multivariate modeling on gene bins, proportion of samples with aberration, difference in proportion between male and female groups for amplifications; the same repeated for deletions and the CNA profile heatmap. The columns represent genes ordered by chromosome. The rows of the heatmap represent samples with the covariate bars indicating clinical variable included this tumour type-specific model.

Supplementary Figure S9. Sex-biased CNAs in lung adenocarcinoma. Each plot shows, from top to bottom: the q-value showing significance of sex from multivariate modeling on gene bins, proportion of samples with aberration, difference in proportion between male and female groups for amplifications; the same repeated for deletions and the CNA profile heatmap. The columns represent genes ordered by chromosome. The rows of the heatmap represent samples with the covariate bars indicating clinical variable included this tumour type-specific model.

Supplementary Figure S10. Sex-biased CNAs in lung squamous cell carcinoma. Each plot shows, from top to bottom: the q-value showing significance of sex from multivariate modeling on gene bins, proportion of samples with aberration, difference in proportion between male and female groups for amplifications; the same repeated for deletions and the CNA profile heatmap. The columns represent genes ordered by chromosome. The rows of the heatmap represent samples with the covariate bars indicating clinical variable included this tumour type-specific model.

Supplementary Figure S11. Pathway analysis for genes in sex-biased copy number aberrations. Network shows terms enriched in genes with sex-biased CNAs. Each node represents a pathway term with node size indicating number of genes in pathway. Tumour types are indicated in colour of nodes. Edges between nodes show gene overlap between pathways.

Supplementary Figure S12. Sex-biased CNAs spanning in KIRC are associated with sex-biased transcriptional changes. Zoom in views of copy number loss comparisons for (a) chromosome 3, (b) chromosome 6 and (c) . Each plot shows, from top to bottom: the multivariate q-value showing significance of sex on copy number loss status on gene bins; difference in proportion between male and female groups for each bin; difference in proportion between male and female groups for each gene; and coefficient estimates of the copy number loss and the sex-copy number loss interaction term in linear regression models for mRNA changes. Darker blue dots indicate genes where copy number loss is associated with significant changes in mRNA (q<0.1). Red dots indicate genes with a significant interaction between sex and copy number (q<0.1), with red lines highlighting genes with greater magnitude of interaction (absolute value of coefficient > 1). Grey lines show centrosomes.

Supplementary Figure S13. Sex-biased CNAs in KIRC are associated with sex- biased transcriptional changes. Female and male adjusted mRNA foldchange compared between copy number loss and no copy number loss from mRNA linear regression modeling. Blue points are genes with decreased mRNA abundance with loss of that gene in male samples (q < 0.05). Pink points are genes with similarly decreased mRNA abundance in female samples (q < 0.05). Red points are genes with significant interactions between copy number loss and sex (q < 0.05).

Supplementary Figure S14. Divergent clinical associations for sex-biased CNAs in kidney clear cell cancer. Female and male survival for UBAC1, which is prognostic in female-derived samples but not male.

Supplementary Figure S15. Functional sex differences in CNAs are associated with outcome in kidney papillary cell carcinoma. (a) Transcriptome differences between the sexes are seen in the mRNA abundance fold-change differences when comparing loss vs. no loss samples in HNSC for each sex. Only genes in sex-biased CNAs were used. Blue points are genes with decreased mRNA in male-derived samples only, pink points are those with decreased mRNA in female-derived samples only and black points are those with similar effects in both sexes. Red points are genes with significant interaction between copy number loss status and sex. (b) Sex-biased mRNA are associated with differential overall survival outcomes between the sexes. Comparison of hazard ratios from Cox proportional hazard models for mRNA abundances in KIRP between the sexes. Two genes have significant association between copy number loss and survival in survival modeling: (c) LCMT1 and (d) C16orf45 are prognostic in women but not men.

Supplementary Figure S16. Sex-biased SNV in kidney papillary carcinoma. Each plot shows, from top to bottom: the q-value showing significance of sex from multivariate modeling, proportion of samples with SNV in gene, difference in proportion between male- and female groups; mutation frequency in all samples; SNV mutation profile across male and female-derived tumours. The covariate bars indicate clinical variable included this tumour type-specific model.

Supplementary Figure S17. The transcriptional impact of sex-biased SNVs in LIHC. BAP1 mutation is associated with decreased BAP1 mRNA abundance in female-derived samples, but not male. CTNNB1 mutation is associated with increased mRNA abundance in both sexes.

Supplementary Figure S18. The transcriptional impact of sex-biased SNVs in STES. RNF213 mutation and sex have a trending interaction effect on RNF213 mRNA abundance.

Supplementary Figure S19. EP400 mutation has a sex-biased effect on SNV mutation prevalence. (A) Mutated EP400 is associated with higher SNV burden. (B) Mutated EP400 also interacts with sex to result in higher SNV burden in female- than in male-derived samples.

Supplementary Figure S20. Sex differences in univariate Cox proportional hazards modeling for NSCLC. (a) FBXO46 and (b) KLF6 abundance are prognostic in males but not females. Conversely, (c) ADCY1 and (d) SPINK1 are prognostic in females but not males.

Supplementary Figure S21. Divergent clinical associations for gene abundance in colon cancer. (a) RPL37A is prognostic in females but not males. (b) SRGAP1 is prognostic in males but not females. (c) ACTL7B and (d) TRRAP are prognostic in both sexes, but in opposite directions. High ACTL7B abundance is a marker of improved prognosis in females and worse prognosis in males. High TRRAP abundance is a marker of improved prognosis in males and worse prognosis in females.

Supplementary Figure S22. Prognostic gene signature performance on combined sex cohort for NSCLC. (a) Receiver operating characteristic curve and (b) Kaplan-Meier analysis of a 100-gene signature fit on the combined sex training cohort and tested on the combined test cohort. The top 100 significant genes from univariate Cox proportional hazards analysis were used.

Supplementary Figure S23. Sex-specific null distributions from random prognostic gene signatures. Randomly generated 100-gene signatures were fit using a combined sex training cohort and tested on female and male test cohorts separately.