Patient-partnered multiomics reveals the molecular architecture of angiosarcoma

Figures

• 

  Fig. 1. Overview of the clinical attributes of patients in the ASCProject cohort. a Schematic of the Angiosarcoma Project and the data types collected. b Patient-reported data of angiosarcoma primary sites. AS Angiosarcoma, RAAS Radiation-associated angiosarcoma, LAAS Lymphedema-associated angiosarcoma. HNFS: Head, Neck, Face, and Scalp. NOS: Not Otherwise Specified. c Prior cancer history of the patients in the study. d Age of angiosarcoma diagnosis stratified by primary sites (n = 252). P-value is calculated from a two-sided Wilcoxon signed-rank test. e Frequency of metastatic events by anatomical sites f Frequency of metastatic co-occurrence by anatomical sites. g Top 15 most commonly prescribed treatments extracted from medical records across different treatment modalities.

• 

  Fig. 2. Transcriptomic landscape of 122 angiosarcoma samples. a–c First 2 principal components of normalized bulk RNA expression across all angiosarcoma samples based on the top 2000 most highly variable genes, colored by a angiosarcoma tumor primary sites, b cutaneous status, and c unsupervised hierarchical clustering. d Heatmap of samples by normalized gene expression highlighting the top 5 overexpressed genes in the most significant MSigDB hallmark pathway and oncogenic pathway per anatomic site. Gene significance and log2FC (FC Fold Change) are based on the value from the most highly expressed site. RAAS Radiation-Associated Angiosarcoma, LAAS Lymphedema-Associated Angiosarcoma, Repr. Representative, TMB Tumor Mutational Burden, log2FC log-2 Fold Change, adj. P: adjusted-p-value, HNFS Head, Neck, Face, Scalp Angiosarcoma, mut/MB Mutation per megabases.

• 

  Fig. 3. Mutational Landscape of 94 angiosarcoma samples. a Comparison of tumor mutational burden of angiosarcomas from different primary sites (n = 94). P-values are calculated from two-sided Wilcoxon signed-rank test. b MutSig2 significance of recurrently mutated genes. c Oncoplot of the top 20 most frequently altered genes in angiosarcoma. Copy number alterations were filtered out if they co-occur with another copy number alteration within 1MB of an OncoKB annotated gene. d Comparison of gene mutation frequency in cutaneous vs non-cutaneous angiosarcoma samples. Padj: adjusted-p values. e Missense mutations found in more than 1 patient; asterisk indicates the missense mutation is predicted to be Pathogenic by AlphaMissense.

• 

  Fig. 4. Pathogenic germline variants in 229 angiosarcoma patients. a All detected putative germline pathogenic variants within the 107 sarcoma genes across the cohort. b Odds ratio of germline pathogenic variant carrier status in angiosarcoma cohort against 21,639 cancer-free controls. The p-values are derived from Wald z-tests on the logistic regression coefficients and are adjusted for multiple testing via the Benjamin–Hochberg procedure. c Distribution of somatic mutations in POT1 and germline pathogenic variants in POT1 (splicing variants not shown).

• 

  Fig. 5. Integrative analysis of POT1 mutations in germline and somatic context. a Comparison of pathogenic variant frequency in gene versus somatic mutation in gene b Genes ranked by the total number of patients with both germline pathogenic variant and somatic mutation in the same gene. c Comparison of angiosarcoma age of onset by POT1 mutation status (n = 250). P-values are calculated from two-sided Wilcoxon signed-rank tests. d Schematic of sample subset for which joint RNA-WES analysis is performed. e Differential expression between tumors carrying POT1 mutation versus POT1 wildtype tumors. P-values are calculated from two-sided Wilcoxon signed-rank tests and adjusted via the Benjamin–Hochberg procedure. f Hallmark gene set over-representation analysis of nominally overexpressed genes in POT1 mutated tumors. P-values are calculated from Fisher’s exact test and adjusted via the Benjamin–Hochberg procedure.