npj Breast Cancer

Recurrence scores based on a 21-gene assay are clinically useful for predicting prognosis and chemotherapy benefit in postmenopausal node-positive breast cancer patients, but its performance in premenopausal patients is inconsistent. Here, we evaluated Ataraxis Breast RISK (ATX), an AI test that predicts recurrence risk, and compared it with the genomic assay. ATX identified high risk patients misclassified as low risk by the genomic assay and therefore may refine selection of patients for adjuvant chemotherapy.

BackgroundGenomic assays such as Oncotype DX, MammaPrint, and Prosigna have transformed risk stratification and treatment selection in early-stage, estrogen receptor-positive (ER+), HER2-negative breast cancers by enabling more precise prognostication and chemotherapy de-escalation in selected patients. However, their clinical utility is limited in lymph nodes positive disease. A major unmet need is the development of compact, mechanistically grounded biomarkers that extend risk and treatment-response prediction to clinically challenging ER+/HER2- subgroups, including lymph node-positive patients. MethodsBuilding on a cell-of-origin framework, we previously established a 70-gene triple hormone receptor (THR; ER, AR, VDR) signature (THR-70) that reflects luminal epithelial differentiation programs and is prognostic across breast cancer subtypes. Here, we refined this framework using interactome-guided clustering to derive a six-gene cell-of-origin signature (THR-6E: KIF4A, KIF2C, CDC20, FAM64A, TPX2, and LMNB2). We evaluated the prognostic and predictive performance of THR-6E across >7,000 breast cancer cases from multiple independent cohorts, assessed treatment-response prediction using endocrine- and chemotherapy-annotated datasets, and performed independent validation in the I-SPY2 adaptive clinical trial. FindingsTHR-6E robustly stratifies relapse-free survival (RFS) in ER+/HER2- breast cancer independent of tumor grade, proliferation status, and subtype. Hazard ratios for RFS were 2.41 (p<1x10-{superscript 1}), 1.61 (p=4.9x10-), and 1.50 (p=6.2x10-3) for grades 1, 2, and 3, respectively, and 2.16 and 1.33 for Luminal A and Luminal B subtypes. THR-6E maintained predictive value across endocrine- and chemotherapy-treated, untreated, lymph node-positive, and lymph node-negative subgroups. Beyond prognosis, THR-6E predicted endocrine therapy response in ER+/HER2-, node-negative disease and chemotherapy response in ER+/HER2-, node-positive disease, achieving approximately 70% sensitivity and specificity (AUC=0.714, p=3.6x10-), with consistent performance across taxane-, anthracycline-, and FEC-based regimens (AUCs 0.71-0.72). Single-cell transcriptomic and proteomic analyses demonstrated that THR-6E expression is specific to normal and malignant breast glandular epithelium, preserved during transformation, and further enriched in metastatic disease. Consistent with a cell-of-origin program, THR-6E genes were rarely mutated in breast cancer and retained normal tissue-like co-expression patterns. In the I-SPY2 trial, THR-6E achieved robust prediction of pathologic complete response across multiple treatment arms (AUCs 0.72-0.94), with an overall AUC of 0.741. InterpretationThese results support a cell-of-origin-anchored approach to biomarker development and challenge purely tissue-agnostic models of therapeutic response. THR-6E represents a compact, biologically interpretable signature that extends prognostic and predictive assessment to clinically relevant ER+/HER2- subgroups, including lymph node-positive disease. Its mechanistic grounding and stable performance across cohorts position THR-6E as a complementary tool to existing multigene assays, warranting prospective diagnostic accuracy studies to define its clinical utility and impact on treatment decision-making.

BackgroundThe management of residual axillary disease after neoadjuvant therapy (NAT) remains controversial, as current recommendations often treat ypN1 breast cancer as a homogeneous entity despite potential prognostic heterogeneity. Evidence supporting uniform axillary surgical strategies across different levels of residual nodal burden is limited. We investigated whether survival associations related to axillary surgical evaluation differ according to residual nodal burden in ypN1 disease, using an adjuvant cohort to validate a SEER-based proxy for surgical extent. MethodsPatients with 1-3 positive lymph nodes were identified in the SEER database (2000-2022) and stratified into neoadjuvant (NAT; n=30,560) and adjuvant (AT; n=197,586) cohorts. Axillary surgical evaluation was categorized as limited (2-3 examined nodes) or extensive ([&ge;]10 examined nodes). Survival was analyzed using Kaplan-Meier methods and log-logistic accelerated failure-time models, adjusted with inverse probability of treatment weighting. ResultsIn the ypN1 cohort, limited axillary evaluation was not associated with inferior overall survival among patients with a single residual positive node (IPTW-adjusted HR: 1.15, p=0.134; time ratio [TR]: 0.86, p=0.184). In contrast, limited evaluation was associated with worse survival in patients with two positive nodes (HR: 1.70, 95%CI 1.54-1.87; TR: 0.58, 95%CI 0.53-0.64). The findings were similar when using breast cancer-specific survival as the endpoint. ConclusionsSurvival associations related to axillary surgical evaluation after NAT vary according to residual nodal burden. Axillary de-escalation appears feasible in patients with a single residual positive node but cannot be extrapolated to those with multiple residual nodes, underscoring heterogeneity within ypN1 disease.

Invasive lobular breast carcinoma (ILC), the most common special histological subtype of breast cancer, is characterized by nearly universal expression of estrogen receptor alpha (ER) and unique sites of metastases, neither of which is fully recapitulated by genetically engineered mouse models. Using reporter-labeled ILC mouse xenografts, herein we used mammary fat pad, tail vein and intracardiac orthotopic growth to analyze spontaneous and experimental metastasis and gene expression. We observed ER-positive primary tumors with single-file histology and collagen deposition, and spontaneous metastasis from the mammary fat pad to bones, ovaries, and brain including the leptomeninges, thereby closely mirroring the growth and metastatic spread of human ILC. Brain metastases showed strong ER staining, confirmed by sequencing analyses which identified estrogen signaling as top activated pathway, and the lesions exhibited robust response to endocrine therapy. In summary, we report endocrine responsive mammary fat pad, tail vein and intracardiac xenografts that faithfully demonstrate unique ILC features and can serve as invaluable pre-clinical translational platforms for validating candidate ILC genetic drivers and testing novel therapeutics.

Nottingham histologic grading is essential for breast cancer prognostication but suffers from inter-observer variability in assessing mitotic activity, nuclear pleomorphism, and tubule formation. We developed MOSAIC (Mammary Oncology Spatial Analysis and Intelligent Classification), an explainable AI framework designed to perform component-wise grading by independently modeling these three histologic features. Model outputs were calibrated using a two-phase pathology study to establish clinically reproducible scoring thresholds and were subsequently evaluated across public datasets and multi-institutional Indian cohorts. MOSAIC demonstrated robust performance, with AI-derived grades providing independent prognostic information (HR >= 1.8 in two datasets, p = < 0.001) and improved survival stratification compared to traditional methods. In pathologist calibration studies, AI-assisted scoring significantly reduced variability, specifically achieving near-perfect agreement in mitotic scoring with a weighted {kappa} up to 0.98. Accuracy and Cohens kappa ({kappa}) analysis further characterized the models technical performance across components: Tubule formation showed the highest agreement (Accuracy >= 0.6607, {kappa} = 0.549), followed by overall Grade (Accuracy = 0.5637, {kappa} = 0.539) and Mitotic activity (Accuracy = 0.4985, {kappa} = 0.4), while Nuclear pleomorphism proved the most challenging (Accuracy = 0.3303, {kappa} = 0.271). Comparative survival models confirmed that AI-derived grades were more significant predictors of risk than manual pathologist-assigned grades, with the AI model yielding a superior global p-value (5.9 x 10-7) and lower AIC (769.61). These results indicate that MOSAIC enables reproducible, interpretable grading by decomposing assessment into pathology-aligned components. By enhancing consistency while preserving prognostic relevance, this framework supports explainable AI as a viable assistive tool for routine breast cancer pathology.

Patient selection and enrolment into phase III randomized clinical trials (RCTs) of adjuvant cyclin-dependent kinase 4 and 6 (CDK4/6) inhibitor therapies depend on accurate risk definition. However, standard clinicopathologic criteria incompletely capture recurrence risk, limiting their efficacy in treatment selection. To assess whether artificial intelligence (AI)-enhanced prognostication may enrich the clinical risk groups utilized in the adjuvant NATALEE trial, we evaluated Ataraxis Breast RISK (ATX), a multimodal AI test that integrates clinical data with morphological features from H&E-stained slides. ATX risk scores were generated for 2,228 patients with HR+/HER2- early breast cancer, of which 918 (41%) were classified as clinical high-risk and 1,310 (59%) were clinical low-risk. ATX was significantly associated with recurrence-free interval in both clinical risk groups and identified high-risk patients not captured by current clinical criteria, as well as individuals with limited benefit despite clinical high-risk classification. Consequently, integration of AI-enhanced risk assessment may improve selection of patients likely to benefit from adjuvant CDK4/6 inhibitors relative to current criteria.

BackgroundEarly and accurate prediction of pathological complete response (pCR) is essential for personalizing neoadjuvant chemotherapy (NACT) in invasive breast cancer. However, most high-performing predictive models rely on costly, multi-modal data that are not routinely available in standard clinical practice. PurposeTo develop and validate Breast Cancer Biological Multi-modal Information Transfer for Response Prediction Model (BC-BioMIXER), a biologically informed predictive model that transfers multi-omics-derived knowledge to routine clinical data, enabling accurate prediction of pathological complete response prior to neoadjuvant chemotherapy initiation. Material and MethodsBC-BioMIXER was developed in a multi-modality cohort of 648 patients with invasive breast cancer (T2-4, any N, M0) incorporating transcriptomic, proteomic, MRI, and clinical data. The model was externally validated in three independent cohorts (total N = 830), including one multi-modality cohort, one clinical trial cohort, and one contemporary real-world cohort. All patients received NACT followed by surgery. The framework employs a teacher-student knowledge-transfer paradigm in which a multi-omics teacher model learns biologically integrated representations that are subsequently transferred to a student model using only routine clinical data. Predictive performance for pCR was benchmarked against a multi-modality reference model and evaluated across cohorts, receptor-defined subgroups (HER2 and hormone receptor [HR]), and treatment groups (NACT with or without immune checkpoint inhibitors [ICI]). Prognostic value was assessed using distant recurrence-free survival (DRFS). The potential to inform immunotherapy decision-making was explored by comparing DRFS between NACT + ICI and NACT-alone groups within model-predicted pCR and non-pCR subgroups. ResultsBC-BioMIXER achieved pCR prediction performance comparable to the multi-modality benchmark (AUC 0.82 vs. 0.85; p = 0.271) and demonstrated consistent discrimination across all validation cohorts (AUCs 0.82, 0.81, and 0.80; all p < 0.001). Patients predicted to achieve pCR experienced significantly improved 3-year DRFS (HR = 0.36; 95% CI, 0.20-0.67; p < 0.001). In patients treated with NACT + ICI, BC-BioMIXER showed numerically superior pCR prediction compared with PD-L1 expression alone (AUC 0.84 vs. 0.72; p = 0.08). Notably, within the model-predicted non-pCR subgroup, patients receiving NACT + ICI had significantly inferior DRFS compared with those receiving NACT alone (HR = 2.70; p = 0.032), whereas no significant difference was observed in the predicted pCR subgroup. ConclusionBC-BioMIXER translates multi-omics-derived biological knowledge into a robust, routine-data-based predictive tool for breast cancer NACT. Its consistent validation across evolving clinical settings and its potential to inform personalized immunotherapy strategies highlight a step toward scalable and accessible precision oncology. HighlightsO_LIBrings multi-omics power to routine clinical practice: Through cross-modality knowledge transfer, BC-BioMIXER leverages transcriptomic and proteomic data during training to enable highly accurate pCR prediction using only standard MRI and clinical variables (AUC 0.82 vs. 0.85 for full multi-modality benchmark, p=0.271). C_LIO_LIConsistently strong and generalizable performance: Validated in three independent cohorts (total N=830), the model maintained robust pCR discrimination (AUC 0.80-0.82, all p<0.001) across receptor subtypes (HR/HER2) and treatment regimens, including with or without immune checkpoint inhibitors. C_LIO_LIGuides personalized immunotherapy de-escalation: In HER2-negative patients predicted as non-pCR, adding ICI to neoadjuvant chemotherapy was associated with significantly worse distant recurrence-free survival (HR 2.70, p=0.032) compared to chemotherapy alone. This effect was not seen in the predicted pCR group, suggesting the model may help identify patients unlikely to benefit from additional immunotherapy. C_LI

Stromal tumor-infiltrating lymphocytes (sTILs) are promising biomarkers for predicting therapeutic outcomes in triple-negative breast cancer (TNBC), with higher sTIL levels correlating with improved chemotherapy response and survival outcomes. Currently, sTILs are manually evaluated by pathologists, which is prone to inter-reader variability. In this study, we have developed an AI-driven TIL segmentation pipeline to process entire diagnostic hematoxylin-and-eosin-stained whole slide images for reproducible scoring (global TILseg scoring) and reliable prognostication. This pipeline was optimized and tested using two independent TNBC patient cohorts (n = 57 in the discovery cohort, n = 43 in the validation cohort) with clinical outcomes and follow-up data. The global scores generated by TILseg showed moderate to high concordance with expert scoring (Spearman R = 0.84-0.89) and improved patient stratification (p-value = 0.0191) as compared to manual scoring (p-value = 0.0663). Additionally, we investigate how the spatial localization of sTILs (spatial TILseg) impact survival outcomes by identifying TILs in selected stromal subsets (0.02-2 mm from the epithelial clusters). Our findings have shown that TILs up to 50 {micro}m from epithelial regions prove to be most prognostic in predicting recurrence-free survival post-neoadjuvant chemotherapy with higher statistical significance than both manual and global TILseg scoring. Further, spatial TILseg scoring was more significantly associated with pathological complete response status in both patient cohorts. In summary, we present an AI-based digital tool for robust sTIL scoring and spatial mapping to enhance its potential as both a diagnostic and prognostic biomarker, particularly in TNBC patients. SIGNIFICANCEAn automated and spatially resolved AI tool for sTILs scoring enhances patient risk stratification based on both response to treatment and recurrence-free survival, establishing its relevance as an independent prognostic marker.

BackgroundThe tumour microenvironment (TME) influences breast cancer progression and treatment response. We investigated whether TME composition predicts tamoxifen benefit in postmenopausal women with oestrogen receptor-positive, HER2-negative (ER+HER2-) breast cancer. MethodsThis study included 513 patients from the Stockholm Tamoxifen (STO-3) trial, which randomised postmenopausal, lymph node-negative women to tamoxifen or no endocrine therapy. Bulk tumour transcriptomes were deconvoluted with the ConsensusTME algorithm to estimate the relative abundance of 18 immune and stromal cell types. A summary score of combined immune cells was created on a per patient basis and evaluated alongside fibroblast and endothelial stromal compartments. Patients were categorised into immune and stromal tertiles on the basis of these scores. Associations between TME composition and tumour characteristics were evaluated using Spearman correlations and Fishers exact test. Tamoxifen benefit was analysed by univariable Kaplan-Meier (log-rank) and multivariable Cox proportional hazards adjusting for age, tumour size, grade, progesterone receptor, Ki-67, and radiotherapy. Differential expression was assessed with limma and pathway enrichment with fgsea using Hallmark gene sets from MSigDB. ResultsLow immune abundance was significantly associated with higher ER expression (Fishers exact test p < 0.001). Among tamoxifen-treated patients, those with low immune scores showed improved distant recurrence-free interval (DRFI) relative to untreated patients (log-rank p < 0.001). Similarly, intermediate endothelial (p < 0.001) and low/intermediate fibroblast abundances (p = 0.042, p = 0.009) were associated with favourable DRFI. In multivariable models, low immune (aHR = 0.17, 95% CI 0.08-0.40), intermediate endothelial (aHR = 0.21, 95% CI 0.09-0.51), and low/intermediate fibroblast tertiles (aHR = 0.50, 95% CI 0.27-0.93; aHR = 0.36, 95% CI 0.17-0.77) retained significance. Transcriptomic analysis revealed enrichment of oestrogen-response, MYC-target, and oxidative-phosphorylation pathways in low-immune and low-fibroblast tumours, while interferon-{gamma} response and allograft rejection pathways were downregulated. ConclusionsTME composition modulates tamoxifen benefit in postmenopausal ER+HER2-breast cancer. Low immune, intermediate endothelial, and low/intermediate fibroblast abundances are associated with improved benefit from tamoxifen, suggesting that both immune and stromal compartments influence endocrine treatment efficacy.

The development and validation of prognostic and predictive biomarkers in breast cancer is limited by the availability of well-annotated datasets linking tumor molecular features to treatment response and survival outcomes. To address this need, we generated an extensive mouse models dataset comprised of 26 immunocompetent mammary tumor models spanning diverse genetic backgrounds, epithelial-mesenchymal states, the basal-luminal axis, and distinct immune microenvironments. For each model, we measured survival under no treatment, immune checkpoint inhibition (ICI), and carboplatin/paclitaxel chemotherapy. We performed RNA-seq on baseline tumors and on 7-day on-treatment samples for both regimens. Using baseline murine tumor gene expression features, we trained a machine learning Elastic Net model that predicted survival outcomes on multiple human breast cancer datasets with performance comparable to that of existing prognostic assays. We next trained models for ICI benefit, using either the untreated or 7-day ICI treated samples; both models predicted ICI benefit on human ICI treated datasets, with the 7-day treated tumor model showing better performance. We also developed a predictor of carboplatin/paclitaxel response that performed well in mice but did not generalize to human chemotherapy cohorts. Finally, we compared multiple computational approaches, including XGBoost, random forests, and support vector regression; all methods successfully predicted survival outcomes, with Elastic Net offering the best performance and interpretability. These results indicate conserved cancer biology between mouse and human tumors for prognosis and ICI response and establish this large preclinical dataset with linked phenotypic and genomic data, as a resource for benchmarking computational methods for survival prediction. SignificanceThe development of a genomically and phenotypically diverse murine tumor dataset with linked treatment outcomes establishes a robust translational resource to develop, test, and benchmark clinically relevant prognostic and therapeutic response biomarkers.

Breast cancer is the most common cancer diagnosis in women. Clinical studies with triple-negative breast cancer (TNBC) are encouraging for immunotherapy combined with chemotherapy (anti-PD-1 with paclitaxel and/or carboplatin). However, additional clinical advances may be pursued more rapidly with assistance from preclinical TNBC models including syngeneic mammary tumor cell lines. Here, we report two mammary tumor cell lines that exhibit differential responsiveness to immunotherapy in vivo. Spontaneous mammary tumors from C57BL/6J MMTV-Cre Trp53fl/+ animals were passaged serially in cell culture and in vivo in the mammary fat pad of fully wildtype animals. The resulting lines, MM001i and MM008i, lost Trp53 and formed 1000 mm3 tumors in the mammary fat pad within 21-28 days. Despite originating from the same genetic background, these lines exhibit differential responses to immunotherapy. For anti-PD-1 therapy, MM001i is poorly responsive and MM008i is strongly responsive with near-complete tumor regression. In comparison, both MM001i and MM008i respond rapidly to anti-CTLA-4 therapy. Both models express unique tumor antigens as evidenced by immunity to subsequent engraftments. Primary MM008i tumors exhibit greater T cell infiltration, and CD8-positive T lymphocytes are required for anti-PD-1 responses. These TNBC models are promising for further mechanistic studies and testing future single and combinatorial therapies.

BackgroundThe genetic architecture of Breast Cancer (BC) in Arab populations remains largely understudied, limiting the precision of current prevention and screening programs. The Emirati Genome Program (EGP), one of the worlds first nation-wide sequencing initiatives, offers an unprecedented opportunity to delineate inherited BC risk across an entire population. MethodsWe analyzed 436,780 EGP individuals, including 229,309 women, integrating whole-genome sequencing (WGS) with electronic health records (EHRs). We quantified the prevalence and penetrance of pathogenic and likely pathogenic (P/LP) variants across 13 NCCN-recommended BC genes, evaluated the performance of established polygenic risk scores (PRS), and reconstructed >48,000 pedigrees to measure familial aggregation. ResultsP/LP variants were identified in 0.84% of women, accounting for 5.2% of BC cases (mean age of 45.9{+/-}11.1 years). Highly penetrant BRCA1 c.4065_4068del (p.Asn1355fs) and BRCA2 c.2808_2811del (p.Ala938Profs) variants showed age-specific cumulative risks of 37.6% and 31% by age 60, respectively, and allele frequencies up to tenfold higher in the Emirati population than in global reference datasets. The European-derived PRS model (PGS000004) demonstrated strong performance, advancing 10-year BC risk onset by a decade for women in the top decile. Family-based PRS discriminated affected from unaffected individuals, revealing higher polygenic risk even within sister pairs. Integration of monogenic, polygenic, and familial data defined a national framework for risk stratification, identifying disease-free women potentially eligible for targeted prevention. ConclusionsNation-scale genome sequencing reveals, for the first time, the comprehensive landscape of inherited BC susceptibility within a Middle Eastern population. The integration of monogenic, polygenic, and familial data establishes a national framework for genomic risk stratification--transforming population genomics into a foundation for precision prevention and early detection in the UAE and beyond.

BackgroundThe breast tumor microenvironment (TME) is a complex milieu composed of many factors contributing to breast cancer (BC) heterogeneity and therapeutic resistance. Aberrant tumor vasculature in the TME limits nutrient and drug delivery, inhibits anti-tumor immunity, and contributes to a lack of cancer therapy efficacy. Utilizing publicly available scRNA-seq datasets, this study characterizes differences between normal breast and breast tumor endothelial cells (EC), provides insights into tumor endothelial cell subtypes, endothelial anergy, and identifies novel, tumor-specific vascular therapeutic targets. MethodsGene expression data from normal and breast tumor tissue samples were integrated, and the EC subset was extracted via canonical gene marker expression. The EC subset was clustered and evaluated for cell subtypes and differentially expressed genes (DEG). Normal EC (NEC) and tumor EC (TEC) markers were further assessed for correlation to bulk gene expression and patient survival outcomes in cBioPortal and Kaplan-Meier Plotter. Cell type gene expression specificity was evaluated in the 3CA single-cell RNA-seq datasets across multiple cancers. ResultsThis analysis revealed differences in NEC and TEC subtype populations. Breast NEC contained similar proportions of venule and capillary populations, while breast TEC demonstrated a majority of the venule subtype. Further, TEC venules were phenotypically distinct from the NEC venules. Consistent with endothelial anergy, suppression of the key adhesion protein SELE was noted, as well as several pro-inflammatory cytokines including IL6, CCL2, and CXCL8, likely downstream of aberrant NF-kB signaling. Differential gene expression analysis identified several TEC specific up-regulated genes compared to NEC, including CLEC14a, IGFBP4, EMCN, and ADM5. CLEC14a, EMCN, and ADM5 were further validated in the single-cell Curated Cancer Cell Atlas (3CA) to be highly specific to the endothelial cell clusters across multiple tumor types, while IGFBP4 was diversely expressed in endothelial, fibroblast, and some malignant cell types. ADM5, a novel tumor vascular marker, was enhanced in TEC venules and less so in arteriole or capillaries. High expression of ADM5 was associated with poor breast cancer patient survival in the basal PAM50 cancer subtype compared to normal and luminal subtypes. Further, across multiple cancer types, high ADM5 expression was associated with reduced patient survival in anti-PD1- and anti-CTLA4-treated patients but not in anti-PDL-treated patients. ConclusionsIntegration of single-cell RNA-seq data identified an anergic-like response in breast TEC and multiple, highly specific markers to TEC not found in normal breast tissue. CLEC14a and EMCN were validated as TEC markers, extending their annotation in breast TEC, and ADM5 identified as a novel TEC marker in breast and other cancers. Moreover, as ADM5 is associated with reduced patient overall survival, this data suggests that a better understanding of ADM5 and other TEC-specific response pathways may provide novel approaches to reactivate anergic TECs and lead to effective therapeutic interventions for cancer patients. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=96 SRC="FIGDIR/small/661087v2_ufig1.gif" ALT="Figure 1"> View larger version (32K): org.highwire.dtl.DTLVardef@a81bf2org.highwire.dtl.DTLVardef@c2b983org.highwire.dtl.DTLVardef@216ab9org.highwire.dtl.DTLVardef@1e5bebb_HPS_FORMAT_FIGEXP M_FIG Graphical Abstract C_FIG

Clinical interpretation of breast cancer sequencing is constrained not by a lack of data but by the absence of an organising framework that translates constellations of co-occurring mutations and copy-number alterations into tumour-level biology with prognostic and therapeutic meaning. This challenge is exemplified by PIK3CA, a clinically actionable alteration often treated as a single-label biomarker despite context-dependent associations with outcome. We analysed >5,000 breast tumours across multiple cohorts using integrated multi-omics (somatic mutations, copy-number, transcriptomic, proteomic and phosphoproteomic profiles) and quantified the directionality of downstream molecular consequences of recurrent alterations relative to TP53-associated trends to infer dominant tumour programmes. This revealed a robust functional organisation comprising (i) a canonical proliferative/replicative programme, enriched for cell-cycle, DNA replication and E2F signalling, and encompassing TP53 mutations and most recurrent CNAs, and (ii) a non-canonical signalling/cell-state programme marked by recurrent mutations including PIK3CA, CDH1, GATA3, MAP3K1 and AKT1, with opposing transcriptomic/proteomic directionality, comparatively lower proliferative output and a systematic tendency towards mutual exclusivity with TP53, consistent with alternative evolutionary routes. To operationalise these findings for clinical use, we developed T-OMICS (Tiered OMICS Classification System), which layers complementary readouts to deliver a single interpretable tumour profile: Tier 1 provides a continuous genomic-risk backbone via a DNA-anchored prognostic RNA signature capturing canonical proliferative/replicative output; Tier 2 assigns programme identity based on the dominant genomic context; Tier 3 quantifies within-programme activity along a continuum; and Tier 4 overlays non-redundant modifier mutations that refine phenotype, vulnerabilities and resistance liabilities, supported by orthogonal proteomic/phosphoproteomic pathway signals. In ER+/HER2- disease, T-OMICS resolves the prognostic ambiguity of PIK3CA by showing that "PIK3CA-mutant" is not a single biological entity: in a predominant low-genomic-score context, PIK3CA aligns with buffered luminal biology and favourable outcomes, whereas in high-score contexts--conditioned by TP53 background and modifier events--PIK3CA can mark adverse biology with distinct dependencies not captured by proliferation-centric readouts; notably, low-score PIK3CA tumours with CDH1 co-mutation shift to significantly worse outcomes. Together, these results establish a programme- and state-aware framework that converts sequencing reports into clinically legible tumour biology to support risk calibration, therapeutic prioritisation and evolution-aware sampling decisions from early-stage through metastatic ER+/HER2- breast cancer. Lay SummaryBreast cancer tumours often carry many genetic changes at the same time. While modern sequencing can identify these changes in detail, the results are frequently presented as long lists of mutations and DNA alterations that are difficult to interpret in terms of how a tumour behaves or how it should be treated. A well-known example is the PIK3CA gene: although it can be targeted with specific drugs, studies have reported mixed results on whether PIK3CA mutations are associated with better or worse outcomes, making it challenging to use this information confidently in clinical care. To address this problem, we analysed genomic (DNA-wide), RNA, and protein data from more than 5,000 breast tumours. We found that many common genomic changes cluster into two main biological "programmes" that reflect distinct ways tumours grow and survive. One programme is driven by rapid cell division and DNA replication and includes TP53 mutations and many common DNA copy-number changes; tumours following this programme tend to be more aggressive. The second programme is less focused on rapid growth and is defined by mutations such as PIK3CA, CDH1, GATA3, MAP3K1, and AKT1, which influence signalling and cell identity rather than directly accelerating proliferation. These programmes reflect broader tumour behaviours rather than the effects of single genes. Importantly, mutations in the second programme are usually not found alongside TP53 mutations, suggesting that breast cancers can develop through distinct biological routes--with some tumours following an alternative pathway (not overtly proliferation-dependent) that shapes their behaviour and may influence which treatments are most appropriate. Based on these findings, we developed a practical classification system, T-OMICS, for ER-positive, HER2-negative breast cancer. T-OMICS summarises which biological programme a tumour follows, how active or aggressive it is within that programme, and whether additional mutations are present that may influence treatment response or resistance. Using this framework, we show that PIK3CA mutations most often occur in a biologically buffered context associated with more favourable outcomes, but when they occur in more aggressive tumours--shaped by other key genetic changes--they can signal a higher-risk disease with different treatment needs. These findings indicate that treatment decisions should be based on the tumours overall biological pattern, not just the presence of a single mutation. By placing sequencing results in this broader context, T-OMICS supports more accurate risk assessment, better treatment planning, and more informed decisions about when to intensify therapy, from early-stage through advanced breast cancer. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=134 SRC="FIGDIR/small/26346715v1_ufig1.gif" ALT="Figure 1"> View larger version (38K): org.highwire.dtl.DTLVardef@18ae796org.highwire.dtl.DTLVardef@6a641dorg.highwire.dtl.DTLVardef@d2be98org.highwire.dtl.DTLVardef@1df1074_HPS_FORMAT_FIGEXP M_FIG O_FLOATNOGraphical SummaryC_FLOATNO C_FIG

A recent randomized clinical trial in non-small cell lung cancer1 confirms what numerous observational studies have reported - time-of-day (ToD) may dramatically influence treatment outcomes in cancer patients2-9. In this recent trial median overall survival (OS) decreased from 28 months in the early ToD arm to 16.8 months in the late ToD arm. We raise the concern that clinical trial outcomes may be influenced by seemingly minor biases in treatment time across arms. We also suggest that by measuring or randomizing treatment-time in clinical trials, we may identify beneficial ToD-dependent treatments that would otherwise be overlooked.

Structured AbstractO_ST_ABSPurpose/ObjectivesC_ST_ABSThe role of postmastectomy radiotherapy (PMRT) in patients with pathologic N1 (pN1) breast cancer, including triple-negative breast cancer (TNBC), remains controversial in the era of modern systemic therapy. We evaluated the association between PMRT and recurrence-free survival (RFS) and overall survival (OS) and identified prognostic factors in a contemporary single-institution pN1 cohort. Materials/MethodsWe retrospectively reviewed female patients with pT1-2N1M0 breast cancer treated with mastectomy between 2016 and 2022. RFS and OS were estimated using Kaplan-Meier methods and compared by PMRT status with log-rank testing. Univariable Cox proportional hazards models assessed associations between clinical factors--including tumor laterality, receptor subtype (TNBC vs non-TNBC), nodal burden, and adjuvant therapies--and survival outcomes, with subgroup analyses by PMRT status and receptor subtype. ResultsFifty-seven patients were included; 22 (38.6%) received PMRT. With a median follow-up of 85 months, PMRT was not associated with improved RFS (median 133 vs 120 months; p=0.256) or OS (not reached vs 195 months; p=0.154). Hormone therapy was significantly associated with improved RFS (HR 0.43; p=0.026) and OS (HR 0.13; p=0.003), while having 2-3 positive lymph nodes predicted worse RFS (HR 2.86; p=0.007). No significant differential benefit from PMRT was observed in patients with TNBC or non-TNBC disease. ConclusionsPMRT was not associated with a survival benefit in this pN1 cohort, including patients with TNBC. Interpretation is limited by modest sample size and statistical power. Outcomes appeared driven by tumor biology, nodal burden, and systemic therapy, supporting individualized PMRT decision-making.

Background: Early breast cancer detection remains central to improving clinical outcomes, yet conventional screening pathways, particularly mammography, have recognized limitations in sensitivity, specificity, and performance in dense breast tissue. Circulating microRNAs (miRNAs) have emerged as promising minimally invasive biomarkers, while artificial intelligence and machine learning (AI/ML) offer powerful tools for identifying diagnostically relevant multi-marker patterns within complex biomarker datasets. This systematic review and meta-analysis evaluated the diagnostic performance of AI/ML-based circulating miRNA signatures for early breast cancer detection. Methods: A systematic search of PubMed/MEDLINE, Scopus, and Web of Science Core Collection was conducted from database inception to 31 December 2025. Studies were eligible if they were original human investigations evaluating circulating miRNAs using an AI/ML-based diagnostic model for breast cancer detection and reporting extractable diagnostic performance metrics. Study selection followed PRISMA 2020 and PRISMA-DTA guidance. Methodological quality was assessed using QUADAS 2. Pooled sensitivity and specificity were synthesized using a bivariate random-effects model, and overall diagnostic performance was summarized using a hierarchical summary receiver operating characteristic framework. Results: Seven studies met the inclusion criteria for qualitative synthesis, with eligible studies contributing to the quantitative analysis depending on data availability. Across the pooled analysis, AI/ML-based circulating miRNA models demonstrated good overall diagnostic performance, with a pooled AUC of 0.905 (95% CI: 0.890 to 0.921), pooled sensitivity of 81.3% (95% CI: 76.8% to 85.2%), and pooled specificity of 87.0% (95% CI: 82.4% to 90.7%). Heterogeneity was moderate for AUC (I2 = 42.3%) and sensitivity (I2 = 38.7%) and low for specificity (I2 = 28.4%). Risk-of-bias assessment showed overall low-to-moderate methodological concern, with patient selection representing the most variable domain. Deeks funnel plot asymmetry test showed no significant evidence of publication bias (p = 0.34). Conclusions: AI/ML based circulating miRNA signatures show promising diagnostic accuracy for early breast cancer detection and may have value as non invasive adjunctive tools within imaging supported diagnostic pathways. However, the evidence base remains limited by methodological heterogeneity, variable validation rigor, and the predominance of retrospective case control designs. Prospective, standardized, and externally validated studies are needed before routine clinical implementation can be justified.

BackgroundAlthough DCIS has a relatively low recurrence rate, many patients still receive adjuvant radiotherapy or endocrine therapy, raising concerns about overtreatment. Reliable biomarkers are therefore needed to predict an individual patients risk and guide treatment decisions. Recent studies suggest that the composition of the tumour-associated stroma (TAS) affects progression and outcome, highlighting TAS-derived biomarkers as promising candidates for further investigation. MethodsWe trained AI models for cell and tumour segmentation using whole slide digital pathology images acquired as part of a retrospective cohort study. We investigated the effects of cell density within both the tumour and the TAS to determine how they correlated with recurrence in the ipsilateral breast. ResultsWe found that the concentration of DCIS lesions on the slide and the density of mitotic figures inside the TAS region were significantly associated with recurrence risk. Additionally, we found some predictive value in the lymphocyte and red blood cell densities in different tumour regions. Stromal composition was shown to associate with recurrence risk, and density-based biomarkers were identified and used to cluster patients into phenotypes with significantly different risk profiles. ConclusionOur findings highlight the prognostic relevance of stromal composition in DCIS, and we identify novel density-based biomarkers that can be used to identify patients who are more likely to experience a local recurrence after breast-conserving surgery alone. These results may aid in developing future risk-stratification tools for breast cancer patients, thereby reducing overtreatment and improving patient care.

BackgroundLong non-coding RNAs (lncRNAs) have emerged as key regulators of tumor biology, however, thus far none have translated to cancer therapies. The lncRNA MALAT1 is overexpressed in more than 20 cancers, including breast cancer and has been shown to function via various mechanisms in a context-dependent manner, in 2D cell lines and mouse models. However, its functional role and therapeutic potential have not been evaluated in clinically relevant patient-derived models. MethodsWe investigated the therapeutic potential of a MALAT1-targeting antisense oligonucleotide (ASO) for breast cancer, using clinically relevant 3D human patient-derived organoids (PDOs) and PDO-xenograft (PDO-X) models. We systematically evaluated the efficiency of MALAT1-targeting ASOs using a biobank of 28 PDO models. Using three independent PDO-X models of triple negative breast cancer (TNBC), we targeted MALAT1 in vivo to study its impact on transcription, alternative splicing, stromal remodeling and metastasis. ResultsAcross PDO-X models, MALAT1 depletion reproducibly drove widespread alternative splicing changes across all event types, particularly intron retention events, accompanied by modest gene expression alterations. Differentially spliced transcripts were enriched for targets of shared cancer-associated transcription factors, and MALAT1 knockdown altered the relative abundance of previously unannotated splicing isoforms. Beyond tumor-intrinsic effects, tumor-specific MALAT1 depletion induced a consistent reduction in macrophage-associated gene signatures and reduced lung metastatic burden. ConclusionsOur data define MALAT1s multifaceted role in TNBC, coordinating alternative splicing, transcriptional fine-tuning, tumor-stroma crosstalk, and metastatic progression. Our study provides strong preclinical evidence supporting MALAT1-targeted ASO therapy and establishes PDO-X models as a clinically relevant platform for functional interrogation of TNBC therapies.

Metastatic breast cancer is a global health concern with a persistently low five-year survival rate. Taxane microtubule stabilizers, including docetaxel (DTX), are the standard of care in various treatment protocols. DTX is used both as a single agent and in combination therapies, with a majority of ER+ breast cancer patients ultimately developing chemoresistance. The mechanisms contributing to chemoresistance involving the tumor microenvironment (TME) have not been fully elucidated. Specifically, the role of vascular cells within the TME, particularly pericytes, is understudied, and their role in promoting chemoresistance remains unknown. Inflammatory cytokines such as interleukin 6 (IL-6) are known to drive drug resistance via activation of the pro-survival JAK/STAT pathway. We found that DTX induced IL-6 secretion of pericytes by at least two-fold compared to vehicle-treated controls in vitro. All tested breast cancer cell lines expressed subunits of the IL-6 receptor (IL-6R) complex, indicating their capacity to respond to JAK/STAT signaling. Conditioned media from DTX-treated pericytes activated STAT3 in ER+ breast cancer cells to levels comparable to recombinant IL-6. Pharmacologic blockade of IL-6 signaling with the IL-6R inhibitor, tocilizumab, reduced DTX-induced STAT3 activation in vitro. Furthermore, combined treatment with tocilizumab and DTX synergistically suppressed the growth of zero-passage patient-derived ER+ breast cancer organoids expressing intact IL-6 signaling. Together, our findings suggest that combining DTX with tocilizumab may revert DTX-induced chemoresistance in ER+ breast cancer patients by inhibiting IL-6-mediated activation of the STAT3 pathway.