Background Molecular subtypes of lung adenocarcinoma (LUAD) with varying prognosis and characteristics have been proposed based on one or two-dimensional studies but are not yet implemented into clinical routine. Epigenetic modifications in cancer cells are independent of sequence variants, directly linked to gene and genome regulation, and thus provide important information to guide subclassification efforts. Methods We performed in-depth epigenomic profiling of 95 primary LUAD samples from a Swedish discovery cohort with comprehensive clinicopathological, epigenomic, genomic, transcriptomic, proteomic, and metabolomic data. Additionally, we estimated pure tumor cell methylomes using a computational approach. We subdivided the discovery cohort into four epigenetic subtypes, the epitypes, reflecting distinct tumor cell methylation states. Resulting epitypes were contrasted based on clinicopathological and molecular features, and our main findings were validated in two additional primary tumor cohorts totaling over 700 samples. Results Of the four DNA methylation epitypes, M1-M4, M1 and M4 were associated with the previously proposed mRNA subtypes Terminal Respiratory Unit and Proximal Proliferative, respectively. Epitypes M2 and M3 showed similar mRNA/protein subtype composition but differed with respect to e.g., higher expression of the LUAD histology-associated NAPSA/surfactant metabolism expression metagene in M3. Genes included in this metagene showed lower DNA methylation in M3, counter to a global tendency towards promoter hypermethylation in this epitype. To further delineate tumor intrinsic links between the epigenomic and expression phenotypes, 62 LUAD cell lines classified into the four epitypes were investigated and recapitulated several characteristics from the tumor epitypes, such as methylation and expression pattens of NAPSA/surfactant genes, highlighting epigenetic states as likely drivers or maintainers of broad tumor phenotypes and differentiation states. Conclusions Dissecting LUAD based on combined biological characteristics using multiomics data has deepened our understanding of the heterogeneity in this complex disease and the mechanisms underlying phenotype formation and maintenance. There remains a critical need for large, publicly accessible, well-annotated multiomic LUAD cohorts to support rigorous subtype discovery and validation, particularly those linked to targeted therapy trial outcomes.

The MAPRE3 gene is aberrantly expressed in several cancers. We profiled DNA methylation in tumor tissues from early‐stage non‐small cell lung cancer (NSCLC) patients and assessed associations with overall survival (OS). Significant CpG probes were validated in The Cancer Genome Atlas (TCGA). The methylation level of cg12821679 MAPRE3 showed significant associations with OS in lung squamous cell carcinoma (LUSC) (HR = 0.32, P = 6.55 × 10 −7 ), but it was not observed in lung adenocarcinoma (LUAD). In LUSC, MAPRE3 expression was significantly correlated with cg12821679 MAPRE3 ( r = 0.17, P = 2.96 × 10 −3 ) and potential trans ‐regulated genes were enriched in the Nicotine addiction pathway. Additionally, MAPRE3 expression showed significant associations with OS in both LUAD and LUSC (LUAD: HR low vs high = 2.28, P = 2.40 × 10 −3 ; LUSC: HR low vs high = 1.61, P = 0.0244). The association between smoking cessation and overall survival was significantly modified by MAPRE3 expression (HR interaction = 0.69, P = 0.0282). Smoking cessation improved OS only in patients with high MAPRE3 expression (HR = 0.56, P = 2.82 × 10 −3 ). We conclude MAPRE3 may predict NSCLC prognosis and influence the prognostic benefit of smoking cessation.

DNA methylation deregulation is an essential feature of tumor biology, influencing cancer formation and pathogenesis. Analyses of DNA methylation in bulk cancer specimens are challenging due to the high data dimensionality, noise, and mixture of different cell types. Here, we characterized methylation dynamics in cancer by investigating the variance structure of bulk tumor DNA methylation data through the identification of groups of highly correlated CpGs, termed CpG cassettes. Using triple-negative breast cancer as a model system, our approach identified co-occurring methylation patterns linked to different intrinsic tumor processes and pathways, gene inactivation, and composition of the tumor immune microenvironment. Our framework also demonstrated the presence of high-variance DNA methylation, seemingly not linked to tumor biology, that could be excluded using chromatin accessibility filtering. Together, this work outlines a comprehensive approach to analyze bulk tumor DNA methylation data, combining tumor purity adjustment, functional CpG filtering, stratification by CpG contexts, and identification of highly correlated CpG modules to enhance tumor-intrinsic DNA methylation patterns and our understanding of processes shaping epigenetic tumor evolution.

BACKGROUND:Homologous recombination (HR) deficiency (HRD) is prevalent in ovarian, prostate, and specific subgroups of breast cancer, particularly triple-negative breast cancer (TNBC). Tumor HRD status can be inferred through DNA-based, RNA-based, functional, or image-based approaches. A comprehensive evaluation of the concordance and discordance among HRD prediction methods derived from these different data modalities has been lacking. In the present study, we systematically compared HRD classifications generated by seven distinct methods within a population-representative early-stage TNBC multi-omics cohort and contrasted them to an FDA-approved assay. METHODS:A total of 235 patients from a reported population-based TNBC cohort from southern Sweden profiled by RNA-sequencing, whole genome sequencing, and with available RAD51 foci staining on tissue microarrays and digital whole slide H&E images were included. Seven different HRD classification methods were applied to available data, including sequencing-based (HRDetect and Classifier of HOmologous Recombination Deficiency, CHORD), copy number-based (scarHRD, and copy number signature 17), functional HR (RAD51-FFPE), mRNA-based, and image-based (DeepHRD) methods. Eighteen selected tumors were analyzed with the Myriad myChoice CDx assay for exploratory comparison. Survival analysis was performed using invasive disease-free survival as clinical endpoint in patients treated with adjuvant standard-of-care chemotherapy. RESULTS:Overall, our results revealed substantial concordance across HRD assessment methods, alongside method-specific discordances attributable to differences in data preprocessing and, importantly, to training strategies that insufficiently account for the well-established clinical and molecular heterogeneity within breast cancer. Sequencing-based methods and scarHRD showed the greatest classification agreement, with discordance to some extent explained by aspects of inadequate tumor cell content, sequencing depth, and fundamental data processing steps (like segmentation). Discordance in mRNA- and image-based classifications appeared associated with molecular subtype features, suggesting that training cohort context may impact performance by incorporating signals (e.g., mRNA expression patterns) that are not specific to HRD status. Despite variation in HRD classification agreement, all seven methods displayed approximately similar prognostic performance in the subset of patients treated with adjuvant chemotherapy. CONCLUSIONS:Collectively, our findings underscore the necessity for rigorous optimization of data processing workflows and threshold definitions to ensure consistency, comparability, and reproducibility across HRD classification platforms.

Treatment decisions in lung cancer rely on comprehensive molecular and histological characterization, but the limited amount of tumor tissue available from diagnostic procedures remains a major challenge. In this proof-of-concept study, we investigated whether preserving small bronchoscopic tumor specimens in RNAlater enables integrated molecular and histological profiling. Bronchial forceps biopsies and endobronchial ultrasound-guided transbronchial needle aspiration (EBUS-TBNA) specimens were collected during the diagnostic work-up of patients with suspected lung cancer, preserved in RNAlater, and used for DNA, RNA, and protein extraction. RNA was analyzed using NanoString gene expression profiling and RNA-sequencing to assess treatment-relevant gene fusions, MET exon 14 skipping events, and histological subtypes. Somatic mutations were identified by massively parallel sequencing, and proteomics-based non-small cell lung cancer (NSCLC) subtypes were characterized by mass spectrometry. High-quality DNA, RNA, and proteins were successfully recovered from all tumors. Gene fusion analysis was successful in 44 of 45 cases. Gene expression-based histological classification showed good concordance with the clinical pathology diagnosis, although some discrepancies were observed. Mutation profiling was fully concordant with routine clinical testing, and comprehensive proteomic data were successfully generated. Collectively, these findings demonstrate that RNAlater-preserved bronchoscopic specimens provide material of sufficient quality for integrated multi-omics analyses, enabling comprehensive molecular profiling from a single, small tumor biopsy.

Triple-negative breast cancer (TNBC) exhibits heterogeneous treatment responses, yet molecular subtypes based on predefined biological pathways show limited prognostic value. We introduce tumor-specific total mRNA expression (TmS), a pathway-agnostic deconvolution metric derived from matched RNA/DNA sequencing, as a robust stratification tool. Analyzing 575 TNBC patients across Western and East Asian populations, TmS outperforms established subtypes in predicting chemotherapy outcomes, stratifying patients into high TmS with favorable prognosis and low TmS with poor prognosis. Stromal enrichment with immune exclusion emerges as a universal feature of chemotherapy-resistant low-TmS tumors across all cohorts. Population-specific features distinguish Asian cohorts: high-TmS tumors exhibit cell cycle-driven proliferation programs, and low-TmS tumors display immune dysfunction with memory B cell enrichment and divergent RAS/mitogen-activated protein kinase (MAPK) activation, compared to Western populations. Despite these differences, extracellular matrix organization represents a conserved therapeutic vulnerability in treatment-resistant low-TmS patients. TmS provides a unifying framework for dissecting TNBC heterogeneity and enabling precision therapy across diverse populations.

Abstract How plasma cells (PCs) shape anti-tumor immunity is unclear. We hypothesized that conflicting prognostic associations reflect differences in immune context and PC ontogeny. We identify extrafollicular (EF)-PCs as an antibody-independent checkpoint that aborts priming by disabling the cDC1→CD8 + T-cell axis in tumor-draining lymph nodes (td-LNs). EF-PCs blunt cDC1 activation and CCR7-guided repositioning into T-cell zones, precluding formation of TCF1⁺ stem-like CD8⁺ T-cells. Depleting EF-PCs in vivo restores cDC1 trafficking, expands the stem-like reservoir, increases intratumoral CD8⁺ infiltration, and restrains tumor growth; benefit is lost with CD8 T-cell ablation. Neither serum transfer nor Fcγ receptor blockade reverses tumor control, supporting a non-canonical, antibody-independent mechanism. Across independent triple-negative breast cancer cohorts, we find EF-PC hyperplasia in td-LNs and tumors; and within immune-cold cases, EF-PC burden stratifies poor prognosis and metastatic risk. A cross-species EF-PC signature maps to a conserved PC-state across cancer types that is linked to poor outcome and immune-checkpoint blockade resistance. EF-PCs thus relocate the dominant failure point to td-LNs and offer a tractable upstream target to convert immune-cold tumors into immune-responsive disease.

Purpose: This study evaluates deep learning (DL) using gene expression (GEX) and preoperatively available clinical data (PreopClinic) to predict sentinel lymph node macro-metastasis (SLNM), and explores their potential for guiding axillary surgery de-escalation and supporting prognostic assessment. Experimental Design: We retrospectively included 6,836 clinically node-negative (cN0) T1-T2 patients with invasive breast cancer who underwent primary surgery from the Swedish SCAN-B cohort. Three DL models—a multilayer perceptron, a pathway-informed sparse neural network, and a transformer—were developed using the development set (n=4,625) and evaluated against XGBoost in the independent test set (n=2,211). Results: The Transformer outperformed other methods for GEX modeling and minimized the need for prior gene selection. In the independent test set, the combined PreopClinic+GEX model significantly improved SLNM prediction (ROC AUC 0.693, P<0.001) and better identified low-risk patients who might avoid unnecessary SLNB (reduction rate 27.2% at a sensitivity of 92.1%, P=0.02) compared to the PreopClinic model alone. Notably, across-subtype training outperformed within-subtype training, improving nodal prediction, especially in TNBC (ROC AUC 0.734; 95% CI: 0.644-0.837), achieving a substantial SLNB reduction rate of 51.5% (95% CI: 43.2-59.9%). Importantly, the derived SLNM predictor showed prognostic significance (P=0.039), and provided complementary information to the established prognostic factors in the ER+HER2- patients recommended for SLNB under the 2025 ASCO guidelines. Conclusions: These findings highlight the Transformer's robustness against noise and effectiveness in capturing informative GEX features across scales, suggesting the potential of integrating GEX data and PreopClinic variables to enable further axillary surgical de-escalation, including for patients with tumor characteristics not reflected in current ASCO recommendations.

Introduction Decisions regarding lung cancer treatment rely on comprehensive molecular and histological characterization, however, limited tumor tissue availability poses major challenges for such analysis. This proof-of-concept study investigated whether preserving small bronchoscopic tumor specimens in RNAlater could enable integrated molecular and histological assessment using a multicomponent tool. Methods Bronchoscopic small tumor specimens, including bronchial forceps biopsies and endobronchial ultrasound-guided transbronchial needle aspiration (EBUS-TBNA), were collected during initial diagnostic work-up for suspected lung cancer. After sampling for routine diagnostics, additional research specimens from the same tumor lesion were obtained and preserved in RNAlater. DNA, RNA, and proteins were extracted for further analyses of 45 cases with confirmed non-resectable lung cancer. RNA was analyzed using a NanoString gene expression assay and RNA sequencing to assess treatment-predictive fusion gene status, METex14 skipping events and histological subtypes. Mutation detection was performed using massive parallel sequencing and proteome-based non-small cell lung cancer (NSCLC) subtypes were further evaluated using mass spectrometry. Results High-quality nucleic acids and proteins were successfully retrieved for the entire cohort. Forty-four tumors were successfully analyzed for gene fusions. The histological subtypes retrieved based on gene expression were adequate in relation to the clinical pathologist's diagnosis, although with some discrepancies. Mutation detection demonstrated total concordance with clinical routine testing and comprehensive proteomics data were successfully acquired. Conclusion RNAlater-preserved small samples provide high-quality nucleic acids and proteins for multi-omics analyses, enabling comprehensive tumor profiling from a single small bronchoscopic specimen.

BACKGROUND:Homologous recombination deficiency (HRD) originating from inactivation of genes like BRCA1/BRCA2 is a targetable abnormality common in triple-negative breast cancer (TNBC). In estrogen-receptor (ER)-positive HER2-negative (ERpHER2n) breast cancer (BC), HRD prevalence and clinical impact are unclear. METHODS:We analyzed 502 ERpHER2n tumors from patients recruited via the population-representative Swedish SCAN-B study by whole genome sequencing (WGS), defining mutational signatures-based HRD, as well as matched transcriptional, DNA methylation, clinicopathological, adjuvant treatment, and outcome data. RESULTS:We show that HRD is much less frequent in ERpHER2n BC (8.4%) compared to TNBC, though induced by similar genetic/epigenetic mechanisms acting on mainly BRCA1/BRCA2/RAD51C/PALB2 together, providing a plausible HR-inactivation mechanism for 71.4% of HRD tumors. Our modelled estimate of HRD in Western European/Nordic BC is ~10-13%. HRD tumors were observed across all PAM50 gene expression subtypes with the exception of Luminal A tumors ( < 1%) and did not exhibit a unique, defining transcriptional or DNA methylation profile. While HRD status was not statistically associated with differences in patient outcome for patients treated with combined chemotherapy and endocrine therapy, a nonsignificant trend of poorer outcome for patients with HRD tumors was observed for patients treated with adjuvant endocrine therapy only. CONCLUSIONS:ERpHER2n HRD tumors show features of aggressive disease, but do not display a distinct transcriptional or DNA methylation profile that clearly differentiates them from HR-proficient tumors. Though numbers are limited, we present early evidence that HRD stratification by WGS could impact therapeutic strategies, as HRD BCs trended to poorer outcomes when not treated with chemotherapy.

Supplementary Figure 7 showing immune metagene rank scores for the IM classifier applied to 23 FUSCC_validation tumors with no IM consensus label from the online TNBCtype tool.

Supplementary Figure 6 showing immune cell fraction estimates for samples obtained pre- and post-treatment.

Supplementary Table 1 providing annotations for included samples.

Clinicopathologic and molecular characteristics of included TNBC cohorts.

Triple-negative breast cancer (TNBC) accounts for 10% to 20% of primary breast cancers and often has early relapses and aggressive progression. An activated tumor immune response can be prognostic in patients with treatment-naïve and chemotherapy-treated TNBC and may be assessed using gene expression data. We derived a stand-alone predictor for a proposed immunomodulatory transcriptional TNBC subtype in a training cohort of 235 patients with primary disease based on random forest modeling of RNA sequencing data. Validation in independent TNBC cohorts totaling more than 1,200 patients demonstrated that the classifier recapitulates the immunomodulatory mRNA subtype classification, is associated with elevated immune expression and diversity of T-cell receptor genes, is associated with response to neoadjuvant chemotherapy, and can separate patients into subgroups with better or worse prognosis after adjuvant chemotherapy. The availability of stand-alone classifiers for mRNA-based prediction may further enhance RNA sequencing’s usability in a more routine clinical context and for translational endpoints in clinical trials. Significance: Tumor immune response has prognostic and treatment predictive value in TNBC and can be estimated by, e.g., mRNA profiling. Translating this association into classifications for single patients requires stand-alone predictors. We have developed one such mRNA classifier that could be applied in future clinical contexts and clinical trials.

INTRODUCTION:Many studies have aimed at identifying additional prognostic tools to guide treatment choices and patient surveillance in lung cancer by assessing the expression of individual proteins through immunohistochemistry (IHC) or, more recently, through gene expression-based signatures. As a proof-of-concept, we used a multi-cohort, gene expression-based discovery and validation strategy to identify genes with prognostic potential in lung adenocarcinoma. The clinical applicability of this strategy was further assessed by evaluating a selection of the markers by IHC. MATERIALS AND METHODS:Publicly available gene expression data sets from six microarray-based studies were divided into four discovery and two validation data sets. First, genes associated with overall survival (OS) in all four discovery data sets were identified. The prognostic potential of each identified gene was then assessed in the two validation data sets, and genes associated with OS in both data sets were considered as potential prognostic markers. Finally, IHC for selected potential prognostic markers was performed in two independent and clinically well-characterized lung cancer cohorts. RESULTS AND CONCLUSIONS:The gene expression-based strategy identified 19 genes with correlation to OS in all six data sets. Out of these genes, we selected Ki67, MCM4 and TYMS for further assessment with IHC. Although an independent prognostic ability of the selected markers could not be confirmed by IHC, this proof-of-concept study demonstrates that by employing a gene expression-based discovery and validation strategy, potential prognostic markers can be identified and further assessed by a technique universally applicable in the clinical practice. The concept of studying potential prognostic markers through gene expression-based strategies, with a subsequent evaluation of the clinical utility, warrants further exploration.

Changes in TIME status in paired pre- and posttreatment NAC patients. A, Sankey plot of predicted IM status before treatment and after treatment at surgery (based on RNA-seq data from surgical tissue) for 36 patients with RD and matched tumor specimens from the SCAN-B_validationNAC cohort, followed by a later recorded distant metastasis event (asterisks). B, Rank scores for the immune response metagene for the 36 RD patients with paired pre- and posttreatment samples stratified by their combined pre- and posttreatment IM prediction. Left: immune rank scores based on RNA-seq data from pretreatment biopsies. Right: immune rank scores based on RNA-seq data from posttreatment surgical tissue. Numbers above boxplots correspond to sample sizes. C, Immune cell type scores imputed with CIBERSORTx for cell types that differed between pre- and posttreatment samples for IM-predicted positive tumors before treatment that were considered IM negative after treatment. Unadjusted P values reported from two-sided paired t tests. D, Strategy to identify differentially expressed genes between predicted IM-positive tumors before treatment that changed to IM negative after treatment (pos/neg) vs. those that remained IM positive (pos/pos). E, Importance scores for the 433 genes in the IM predictor with scores > 0. Scores for the eight genes from the set of 51 identified in (D) overlapping with the 433 are marked by points. F, FPKM gene expression (gex) data for the interferon signaling–associated genes CD38 (hallmark IFNγ response) and CXCL10 (hallmark IFNα response) from pretreatment and posttreatment samples stratified by their predicted IM status. Two-sided P values calculated using a Wilcoxon test. Gene expression of each gene in IM-predicted negative tumors both before and after treatment (neg/neg) included for reference. G, Same as (F) but for the immune-inhibitory genes PD-L1 (CD274) and LAG3. DEG, differentially expressed gene; GSEA, gene set enrichment analysis.

Introduction: Tumor-infiltrating lymphocytes (TILs) have become a significant biomarker during recent years, showcasing its predictive and prognostic potential for early and metastatic triple-negative breast cancer (TNBC). However, pathologist-read stromal TILs (sTILs) remain a semi-quantitative biomarker, susceptible to inter-observer variability. With the surge in Artificial Intelligence (AI) research, various automated approaches have been proposed to score TILs with the promise to overcome the limitations of manual assessment. However, there is a lack of studies comparing different AI models in both analytical and clinical validity with respect to mimicking the challenges of clinical practice. Methods: In this study, we aimed to investigate the variability among ten AI-based TILs scoring models (seven own-developed machine learning models in QuPath –KNN, Random Forest, Neural Network– and three pre-trained deep learning models –HoverNet Graham et al. Medical Image Analysis 2019, CellViT Hörst et al. Medical Image Analysis 2024, Abousamra et al. Frontiers in Oncology 2022–) with respect to their analytical and clinical validity on internal and external validation sets. The development cohort consisted of diagnostic tissue slides of 79 women with surgically resected primary invasive TNBC tumors diagnosed between 2012 and 2016 from the Yale School of Medicine. An independent prospective set comprising of 215 TNBC patients from Sweden diagnosed between 2010 and 2015, with 4 years median follow-up, was used for assessing the models’ clinical validity. The gold standard of this study regards manual sTILs scoring from two expert pathologists. Results: Moderate correlation in analytical validity (Internal validation set: Spearman’s r= 0.72-0.84, p<0.001; External validation set: Spearman’s r=0.63-0.73, p<0.001) is demonstrated across AI methodologies and training strategies. Training on progressively increasing number of samples improved the correlation with sTILs in internal (10 patients:r=0.79, 20:r=0.81, 30:r=0.82, 40:r=0.84, 50:r=0.83, p<0.001) but not in the external validation sets (10:r=0.70, 20:r=0.68, 30:r=0.70, 40:r=0.68, 50:r=0.73, p<0.001). HoverNet & CellViT achieved the second highest correlation with sTILs in the internal validation set (r=0.83, p<0.001) but second and third to worst in the external validation set (r=0.67 & r=0.64, p<0.001). Variabilities in the distribution of TILs scores were identified across models. Interestingly, eight out of ten models (KNN, RF, NN and HoverNet), even less extensively trained ones, showed statistically significant prognostic potential, with similar and overlapping hazard ratios (HR) in the external validation cohort (Cox regression based on IDFS-endpoint and dichotomized TILs scores at 10%, HRadjusted=0.38-0.50, p<0.047). For reference, manual sTILs demonstrated a HRadjusted=0.43 (p=0.003). Conclusion: Most AI TIL methods demonstrated similar and statistically significant clinical validity, which we believe may be attributed to the intrinsic robustness of TILs as a biomarker. The analytical discrepancies between the AI models should not be overlooked; rather, we believe that there is a need for a large and diverse clinical benchmark dataset to be used for independent model validation ensuring the comparability and reliability of AI tools before integration into the clinical practice. Citation Format: Nikolaos Tsiknakis, Joan Martinez Vidal, Johan Staaf, Ana Bosch, Anna Ehinger, Emma Nimeus, Roberto Salgado, Yalai Bai, David L. Rimm, Johan Hartman, Balazs Acs. Comparison of analytical and prognostic performance among various Artificial Intelligence models for Tumor Infiltrating Lymphocytes scoring in Triple Negative Breast Cancer: An independent validation on a prospective cohort [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2024; 2024 Dec 10-13; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2025;31(12 Suppl):Abstract nr P4-03-16.

BACKGROUND:Breast cancer is the most frequently diagnosed cancer in women. Survival is generally considered favourable, yet some patients remain at risk of early death. We aimed to assess whether comprehensive whole-genome sequencing (WGS) linked to mortality data could add prognostic value to existing clinical measures and identify patients who might respond to targeted therapeutics. METHODS:In this integrative, retrospective analysis, we analysed 2445 breast cancer tumours (any stage and molecular subtype) collected from 2403 patients recruited through 13 National Health Service Genomic Medicine Centres or hospitals in England affiliated to the 100 000 Genomes Project (100kGP) between 2012 and 2018. We linked 2208 (90%) cases with clinical data; mortality data were obtained for 1188 patients. Following high-depth WGS of tumour and matched normal DNA, we performed comprehensive WGS profiling seeking driver mutations, mutational signatures, and compound algorithmic scores for homologous recombination repair deficiency (HRD), mismatch repair deficiency, and tumour mutational burden. Data from 1803 additional patients with breast cancer from three independent cohorts were used to validate various findings. To evaluate the prognostic value of WGS features, we performed univariable and multivariable Cox regression on data from patients with stage I-III, ER-positive, HER2-negative breast cancer with a cancer-specific mortality endpoint (around 5-year follow-up). FINDINGS:Among 2445 tumours in the 100kGP breast cancer cohort, we observed genomic characteristics with immediate personalised medicine potential in 656 (26·8%), including features reporting HRD (298 [12·2%] total cases and 76 [6·3%] ER-positive, HER2-negative cases), highly individualised driver events, mutations underpinning resistance to endocrine therapy, and mutational signatures indicating therapeutic vulnerabilities. 373 (15·2%) cases had WGS features with potential for translational research, including compromised base excision repair and non-homologous end-joining dependency. Structural variation burden (hazard ratio 3·9 [95 CI% 2·4-6·2]; p<0·0001), high levels of APOBEC signatures (2·5 [1·6-4·1]; p<0·0001), and TP53 drivers (3·9 [2·4-6·2]; p<0·0001) were independently prognostic of customary clinical measures (age at diagnosis, stage, and grade) in patients with ER-positive, HER2-negative breast cancer. We developed a prognosticator for ER-positive, HER2-negative breast cancer capable of identifying patients who require either increased intervention or therapy de-escalation, validating the framework in the independent Swedish Cancerome Analysis Network-Breast (SCAN-B) dataset. INTERPRETATION:We show that breast cancer genomes are rich in predictive and prognostic value. We propose a two-step model for effective clinical application. First, the identification of candidates for targeted therapies or clinical trials using highly individualised genomic markers. Second, for patients without such features, the implementation of enhanced prognostication using genomic features alongside existing clinical decision-making factors. FUNDING:National Institute of Health Research, Breast Cancer Research Foundation, Dr Josef Steiner Cancer Research Award 2019, Basser Gray Prime Award 2020, Cancer Research UK, Sir Jeffrey Cheah Early Career Fellowship, the Mats Paulsson Foundation, the Fru Berta Kamprads Foundation, and the Swedish Research Council.