Supplementary Table 8 shows summary statistics of 16 SNPs associated with 23 concordant dQTLs across cohorts

Supplementary Table 6 displays the number of dQTLs identified for each somatic driver in each analysis strategy. Summary statistics from local dQTL associations. Statistics from logistic regression correcting for five genetic principal components, age and somatic mutation burden. OR = odds ratio; SE = standard error; L95 = lower 95% confidence interval; U95 = upper 95% confidence interval

Supplementary Table 9 summarizes the characterization of 16 SNPs associated with 23 concordant dQTLs.

Supplementary Table 1 shows a feature-by-patient summary matrix. For each patient, this table provides version information of bioinformatics tools, summary sequencing statistics, mutational density metrics, clinical information and driver mutation status for all drivers detected using GISTIC and ActiveDriverWGS as or identified in Armenia and colleagues 2018. Two additional tables are provided for the clinical and mutation data for the discovery and replication cohorts of dQTL discovery.

Supplementary Table 3 illustrates driver co-occurrence analysis, driver clusters, and associations of drivers with clinical features

Supplementary Table 11 reports percentages of cross-individual contamination for each sample and sequencing lane.

Adrenal and extra-adrenal paragangliomas (PPGLs) are highly heritable non-epithelial neuroendocrine neoplasms. Through a retrospective chart review of 110 individuals diagnosed with PPGLs at the University Health Network in Toronto, Canada (2011–2023), we characterized germline findings, tumor features, self-reported ethnicity, and variant reclassification across a multi-ethnic cohort using targeted germline panel sequencing, whole genome sequencing (WGS), and optical genome mapping (OGM). Panel sequencing identified pathogenic or likely pathogenic (P/LP) germline variants in 28.18

Supplementary Table 5 displays summary statistics from PRS and HOXB13 associated with somatic drivers. β and P-value from logistic regression correcting for five genetic principal components, age and somatic mutation burden. FDR = false discovery rate.

Supplementary Table 7 provides summary statistics from distal dQTL associations

Supplementary Table 2 displays results from driver selection, driver groupings and driver associations.

Supplementary Table 10 lists dQTL SNPs identified as eQTLs in prostate tissue in GTEx.

Supplementary Table 4 shows driver selection for dQTL nomination and prevalence of drivers in cohorts

Supplementary Figures & Figure Legends. Supplementary Figure 1 | Cohort Structure and Analysis. Supplementary Figure 2 | CNA Evolution & Transcriptomic Effects. Supplementary Figure 3 | Properties of Driver Mutations. Supplementary Figure 4 | Pathway & Signature Analysis of Driver Genes. Supplementary Figure 5 | Patterns of Mutational Drivers. Supplementary Figure 6 | Molecular Correlates of Clinical Behavior. Supplementary Figure 7 | Heterogeneity of Driver-Clinical Associations. Supplementary Figure 8 | Cohort Characteristics and Risk dQTL Replication. Supplementary Figure 9 | Local dQTLs Discovery. Supplementary Figure 10 | Replication of dQTLs. Supplementary Figure 11 | Enrichment of Sub-threshold dQTLs. Supplementary Figure 12 | Molecular Characterization of dQTLs. Supplementary Figure 13 | Association of dQTL Risk SNPs with eQTL and IMS. Supplementary Figure 14 | Clinical Characterization of dQTLs.

Supplementary Table S2: Sample Summary. List of patients and samples (timepoints) included in the current study.

Objectives: Uterine carcinosarcoma (UCS) is an aggressive malignancy characterized by epithelial (C) and mesenchymal (S) components, with complex biology and poor treatment response. This study aims to enhance understanding of UCS through genomic, epigenomic, and transcriptomic analysis. Methods: Microdissected (C and S) tumor samples were processed for whole-genome sequencing (WGS), RNA-seqencing, and enzymatic methylation sequencing (EM-Seq). Multiplex immunohistochemistry (mIHC) and computational pathology techniques were employed to assess tumour microenvironment (TME). Results: WGS and EM-seq of 18 samples from 9 patients revealed a low tumor mutation burden (TMB; median = 0.97 mutations/Mb) and no evidence of microsatellite instability (MSI). Driver mutations were identified in TP53 (94 %), PIK3CA (33 %), and PPP2R1A (22 %). Copy-number (CN) analysis revealed recurrent amplifications of MYC (67 %), PIK3CA (61 %), CCNE1 (56 %), AKT2 (44 %), and SMARCA4 (39 %). Comparative analysis of the C and S regions revealed no significant differences in mutation frequency, CN, transcriptomic and methylomic profiles. Both regions exhibited global hypomethylation, with functional enrichment for xenobiotic metabolism pathways in C and epithelial-to-mesenchymal transition pathways in S regions. Comparitive mIHC performed on 21 cases showed similar T cell and B cell densities, but a higher density of tumour-associated macrophages and PD-L1+ cells in the S component. Computational morphologic analysis showed substantial histomorphologic heterogeneity within and across UCS cases. Conclusion: By elucidating the complex interplay between the epithelial and mesenchymal components, this study enhances our understanding of UCS and informs the development of novel therapeutic strategies targeting both genomic alterations and the TME.

Newly diagnosed prostate cancers differ dramatically in mutational composition and lethality. The most accurate clinical predictor of lethality is tumor tissue architecture, quantified as tumor grade. To interrogate the evolutionary origins of prostate cancer heterogeneity, we analyzed 666 prostate tumor whole genomes. We identified a compendium of 223 recurrently mutated driver regions, most influencing downstream mutational processes and gene expression. We identified and validated individual germline variants that predispose tumors to acquire specific somatic driver mutations: these explain heterogeneity in disease presentation and ancestry differences. High-grade tumors have a superset of the drivers in lower-grade tumors, including increased frequency of BRCA2 and MYC mutations. Grade-associated driver mutations occur early in tumor evolution, and their earlier occurrence strongly predicts cancer relapse and metastasis. Our data suggest high- and low-grade prostate tumors both emerge from a common premalignant field, influenced by germline genomic context and stochastic mutation timing.Significance: This study uncovered 223 recurrently mutated driver regions using the largest cohort of prostate tumors to date. It reveals associations between germline SNPs, somatic drivers, and tumor aggression, offering significant insights into how prostate tumor evolution is shaped by germline factors and the timing of somatic mutations.

Abstract Newly diagnosed prostate cancers differ dramatically in mutational composition and lethality. The most accurate clinical predictor of lethality is tumor tissue architecture, quantified as tumor grade. To interrogate the evolutionary origins of prostate cancer heterogeneity, we analyzed 666 prostate tumor whole genomes. We identified a compendium of 223 recurrently mutated driver regions, most influencing downstream mutational processes and gene expression. We identified and validated individual germline variants that predispose tumors to acquire specific somatic driver mutations: these explain heterogeneity in disease presentation and ancestry differences. High-grade tumors have a superset of the drivers in lower-grade tumors, including increased frequency of BRCA2 and MYC mutations. Grade-associated driver mutations occur early in tumor evolution, and their earlier occurrence strongly predicts cancer relapse and metastasis. Our data suggest high- and low-grade prostate tumors both emerge from a common premalignant field, influenced by germline genomic context and stochastic mutation timing. Significance: This study uncovered 223 recurrently mutated driver regions using the largest cohort of prostate tumors to date. It reveals associations between germline SNPs, somatic drivers, and tumor aggression, offering significant insights into how prostate tumor evolution is shaped by germline factors and the timing of somatic mutations.

Monitoring minimal residual disease (MRD) is critical in multiple myeloma (MM) to predict outcomes and guide therapy. Traditional bone marrow (BM) aspirates for MRD detection are invasive and limited by sample quality. We therefore explored cell-free DNA (cfDNA) fragmentation as a less invasive alternative for MRD detection. We performed 30-40X whole-genome sequencing (WGS) on peripheral blood cfDNA from 45 MM patients, collecting baseline (n = 45) and follow-up samples (n = 98) from eight Canadian sites (M4 and IMMAGINE studies) and one U.S. site (SPORE study), plus 25 healthy controls. Samples were collected post-induction therapy (n = 13), 100 days post-autologous stem cell transplantation (ACST, n = 36), and during maintenance therapy (n = 49). Multiparameter flow cytometry (MFC) MRD testing was conducted on 77 samples. We evaluated insert size metrics and MM-specific chromatin accessibility (from Ordoñez et al., 2020) using Griffin (described in Doebley et al., 2022). Baseline samples had a higher proportion of short fragments (20-150 bp) than follow-up samples (p = 0.0037) and healthy controls (p = 0.019). A fragment score (FS) based on weighted fragment distribution (per Vessies et al., 2022) was highest at baseline (mean = -0.384), followed by MRD-positive (mean = -0.495) and MRD-negative (mean = -0.550) samples. MRD-negative FS was lower than baseline samples (p < 0.001) but higher than healthy controls (mean = -0.639; p = 0.016). Baseline samples had lower coverage at MM-specific chromatin regions than MRD timepoints (p < 0.05; baseline mean = 0.982; MRD-positive = 0.989; MRD-negative = 0.994; healthy = 0.993), suggesting higher transcriptional activity pre-treatment. MRD-negative samples had higher coverage than MRD-positive cases (p < 0.01). Using logistic regression, we assessed the predictive performance of FS and coverage at MM-specific sites. FS alone achieved an AUC of 0.632 (sensitivity [SN] 39.4%, specificity [SP] 92.1%, accuracy [AC] 67.6%). Coverage yielded an AUC of 0.681 (SN 75.8%, SP 60.5%, AC 67.6%). Combining FS and coverage at MM sites improved performance (AUC = 0.734; SN 48.5%, SP 92.1%, AC 71.8%). We next evaluated whether adjusting the BM tumor cell percentage cutoff by MFC could improve performance. At 0.017% (1 in 5, 789 cells) for MRD positivity, the combined model’s accuracy reached 83% with an AUC of 0.757 (SN 17%, SP 97%). cfDNA fragmentomic analysis achieved high specificity (92-97%) in detecting MRD, providing a less invasive alternative to facilitate serial monitoring. While sensitivity remains an area for improvement, the high specificity of cfDNA makes it valuable for confirming MRD negativity and reducing invasive procedures. Future efforts aim to enhance sensitivity by integrating additional fragmentomic features. Dor D. Abelman, Jenna Eagles, Aimee Wong, Saumil Shah, Stephanie Pedersen, Stephenie Prokopec, David S. Scott, Sarah Bridges, Darrell White, Irwindeep Sandhu, Kevin Song, Esteban Braggio, Alli Murugesan, Anthony Reiman, Suzanne Trudel, Trevor J. Pugh. Tumor-independent monitoring of minimal residual disease in multiple myeloma using cfDNA fragmentomics [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 3250.

Supplementary Table S3: Targeted Sequencing Panel. List of regions included in the CHARM+EVOLVE panel used for targeted sequencing; coordinates are for human reference hg38.

Supplementary Figure 4: Distribution of cfDNA fragment length