系统识别乳腺癌中的串联重复表型（TDP）亚型，解析不同亚型的分子特征及其与预后的关系。

收集来自癌症基因组图谱计划（TCGA）的乳腺癌样本1 098例，来自癌症细胞系百科全书（CCLE）的60株乳腺癌细胞系。基于拷贝数变异（CNV）数据识别串联重复事件并构建TDP评分。根据TDP评分将样本分为TDP组（147例，TDP评分>－0.710且TD事件数≥20个）和非TDP组（409例，TDP评分<－0.835或TD事件数<20）。通过高斯混合模型将TDP样本划分为6种TDP亚型：短片段单峰型（组1）、中片段单峰型（组2）、长片段单峰型（组3）和3种单峰组合的双峰混合型（组1/2、组1/3和组2/3）。进一步将组1、组1/2、组1/3归类为小片段串联重复组（SSG），组2、组3、组2/3归为大片段串联重复组（LSG）。分析不同TDP分型的CNV复杂度、临床特征及预后情况，并进行GO和KEGG功能富集分析和药物敏感性分析。采用Kaplan-Meier方法进行生存分析，组间比较采用Log-rank检验。

TCGA数据库中共识别出147例TDP患者，其中SSG组25例（组1、组1/2、组1/3分别为2、1、22例），LSG组122例（组2、组3、组2/3分别为2、50、70例）。非TDP组（409例）、SSG组和LSG组CNV复杂度分别为7.55（7.20，8.03）、8.47（8.29，8.78）和8.37（7.98，8.65），三组比较，差异有统计学意义（H=135.12，P<0.001）。功能富集分析显示，SSG组更倾向于涉及抑癌基因及DNA损伤修复等通路；组2/3主要表现为致癌基因扩增，并富集于肿瘤相关信号通路。CCLE乳腺癌细胞系中共识别出47株TDP细胞系，其中SSG组40株（组1、组1/2、组1/3分别为2、37、1株），LSG组7株（组2、组3、组2/3分别为0、0、7株）。非TDP组（12株）、SSG组和LSG组CNV复杂度分别为8. 91（8. 76，9. 07）、9. 95（9. 78，10. 28）和9. 82（9. 72，9. 91），三组比较，差异有统计学意义（H=28.86，P<0. 001）。药物敏感性分析显示，LSG组细胞系在17-AAG和紫杉醇处理后的中位IC₅₀值高于SSG组。生存分析显示，SSG组5年OS率为100.0%，LSG组5年OS率为80.5%（95%CI：71.3%~90.9%），组间差异有统计学意义（χ²=4.90，P=0.027）。

本研究基于CNV数据识别出乳腺癌TDP并划分为6种亚型。不同TDP亚型在CNV负荷、驱动基因扩增、功能通路、药物敏感性及预后方面存在差异。其中，LSG表现出致癌基因扩增及较差预后，提示TDP分型可为乳腺癌分子异质性分析和预后分层提供参考。

To systematically identify different subtypes of tandem duplication phenotype (TDP) in breast cancer and analyze their molecular characteristics and prognostic relevance.

A total of 1 098 breast cancer samples from the Cancer Genome Atlas (TCGA) and 60 breast cancer cell lines from the Cancer Cell Line Encyclopedia (CCLE) were collected. Tandem duplication events were identified based on copy number variation (CNV) data, and TDP scores were calculated. Samples were classified into the TDP group (TDP score > −0.710 and number of TD events ≥20) and the non-TDP group (TDP score < −0.835 or number of TD events < 20). TDP samples were classified into six subtypes using a Gaussian mixture model: short-segment unimodal type (group 1), intermediate-segment unimodal type (group 2), long-segment unimodal type (group 3), and three bimodal mixed types composed of unimodal patterns (group 1/2, group 1/3, and group 2/3). Group 1, group 1/2, and group 1/3 were further categorized as the short-segment tandem duplication group (SSG), whereas group 2, group 3, and group 2/3 were categorized as the large-segment tandem duplication group (LSG). The CNV complexity, clinical characteristics, and prognosis of different TDP subtypes were analyzed, along with GO and KEGG functional enrichment analyses and drug sensitivity analysis. Survival analysis was performed using the Kaplan–Meier method, and differences between groups were compared using the Log-rank test.

A total of 147 TDP patients were identified in the TCGA database, including 25 cases in the SSG group (2, 1 and 22 cases in groups 1, 1/2, and 1/3, respectively) and 122 cases in the LSG group (2, 50 and 70 cases in groups 2, 3, and 2/3, respectively). The CNV complexity values in the non-TDP (409 cases), SSG, and LSG groups were 7.55 (7.20, 8.03), 8.47 (8.29, 8.78), and 8.37 (7.98, 8.65), respectively, with a statistically significant difference among the three groups (H=135.12, P<0.001). Functional enrichment analysis showed that the SSG was more likely to involve tumor suppressor genes and pathways such as DNA damage repair, whereas group 2/3 was mainly characterized by oncogene amplification and enrichment in tumor-related signaling pathways. In the CCLE cohort, 47 TDP strains were identified, including 40 strains in the SSG group (2, 37 and 1 strain in groups 1, 1/2, and 1/3, respectively) and 7 strains in the LSG group (0, 0 and 7 strains in groups 2, 3, and 2/3, respectively). The CNV complexity values in the non-TDP (12 strains), SSG, and LSG groups were 8.91 (8.76，9.07), 9.95 (9.78，10.28), and 9.82 (9.72，9.91), respectively, with a statistically significant difference among the three groups (H=28.86, P<0.001). Exploratory drug sensitivity analysis showed that the median IC₅₀ values of LSG cell lines treated with 17-AAG and paclitaxel were higher than those of SSG cell lines. Survival analysis showed that the 5-year OS was 100.0% in the SSG and 80.5% in the LSG (95%CI: 71.3%–90.9%), with a statistically significant difference between 2 groups (χ²=4.90, P=0.027).

This study identified TDP in breast cancer based on CNV data and classified it into six subtypes. Different TDP subtypes showed differences in CNV burden, driver gene amplification, functional pathways, drug sensitivity, and prognosis. LSG was characterized by oncogene amplification and poorer prognosis, suggesting that TDP classification may provide a reference for analyzing molecular heterogeneity and prognostic stratification in breast cancer.