癌症中情境依赖性的功能性非整倍体

Background: Aneuploidy shows strong cancer-type specificity, yet its mechanistic basis remains unclear. Classic events like 8q gain in ovarian cancer or 17p loss in breast cancer highlight recurrent patterns, but the standard 98% arm-coverage rule obscures meaningful CNV boundaries. We hypothesised that recurrent breakpoints across tumours provide a more functional definition of aneuploidy. To address this, we developed BAGEL, which quantifies copy-number alterations by defining CNV segments using BISCUT breakpoints rather than fixed chromosome-arm thresholds. Methods: We mapped recurrent CNV breakpoints across 38 cancer types using BISCUT and validated them across nine datasets spanning multiple sequencing platforms and segmentation algorithms (APOLLO, CPTAC3, CGCI, TARGET, DepMap, PCAWG, TCGA and two in-house cohorts). BAGEL-derived aneuploidy scores were tested for prognostic value using Cox and Kaplan-Meier analyses in TCGA high-grade ovarian cancer, with validation in PCAWG. Functional effects were assessed by integrating DepMap CRISPR screens and modelling essentiality differences using mixed-effects models. Chromatin constraints were quantified using TAD penetration scores from colorectal tumours and breast and lung cancer cell lines. Results: BAGEL replaced the arbitrary 98% arm rule by defining CNV segments directly from shared breakpoints. Breakpoints were highly reproducible (mean deviation = 0.089), including 34 nearly identical sites across eight HGSOC cohorts. Prognostic modelling by multivariate cox showed that Breakpoint-defined aneuploidy strongly predicted survival in both HGSOC and breast invasive carcinoma, with TCGA dataset-built model p<0.0001 for both and showed consistent trend in PCAWG dataset (5 year OS p = 0.085 for HGSOC and p = 0.026 for breast invasive carcinoma). Mixed-effects modelling between 584 cancer-pairs showed significant interaction between cancer type and presence of breakpoint event in 123 arms (FDR < 0.05), indicating that the CRISPR essentiality difference between aneuploid and non-aneuploid samples varies substantially across cancer types. Hi-C-derived TAD penetration scores demonstrated consistent structural constraints across tumour and cell-line datasets. Breakpoints showed a subtle but reproducible bias toward TAD boundaries, particularly for negatively selected events, indicating that 3D genome architecture shapes permissible CNA breakpoint positions. Conclusions: Breakpoint-defined aneuploidy captures recurrent, biologically constrained CNV segments, reveals cancer-type-specific selective pressures, identifies aneuploidy-linked gene-essentiality shifts, and exposes a reproducible positional bias toward TAD boundaries. These findings demonstrate that aneuploidy evolution is shaped jointly by selective fitness pressures and 3D chromatin architecture.