Triple-negative breast cancer, defined by the absence of estrogen receptor and progesterone receptor expression and the lack of HER2 amplification, accounts for roughly 10 to 15 percent of all breast cancers but carries a disproportionate share of early recurrences and deaths. Because the tumor lacks the molecular targets that make other breast cancer subtypes vulnerable to endocrine therapy or HER2-directed drugs, cytotoxic chemotherapy has remained the backbone of treatment. Historically, anthracycline-and-taxane regimens have been the standard of care, producing pathologic complete response rates of roughly 37 to 41 percent. When patients achieve a pathologic complete response, meaning no residual invasive cancer is found in the breast or lymph nodes at the time of surgery, their five-year event-free survival approaches 90 percent, compared with 57 percent for those who do not. That makes the response a powerful surrogate for cure and a critical endpoint for evaluating treatment strategies.

The trouble with anthracyclines lies in what they leave behind. Pooled data from randomized controlled trials show that patients treated with anthracycline-containing regimens face more than a five-fold higher risk of clinical cardiotoxicity, as well as sharply elevated risks of acute myeloid leukemia and myelodysplastic syndrome. Long-term follow-up studies in breast cancer survivors have documented heart failure, arrhythmias, and secondary blood cancers years or decades after the original diagnosis. For young women facing the prospect of decades of survival after successful breast cancer treatment, these late effects are not trivial side effects; they are life-altering or life-shortening complications in their own right. Reducing or eliminating anthracycline exposure without sacrificing cancer control has therefore become one of the most consequential questions in early breast cancer management.

At the same time, the treatment landscape has been transformed by two major developments. The first is the incorporation of carboplatin, a platinum chemotherapy agent, into neoadjuvant regimens. Trials such as GeparSixto, CALGB 40603, and BrighTNess demonstrated that adding carboplatin increases pathologic complete response rates in triple-negative disease, although often at the cost of substantially more hematologic toxicity, treatment delays, and dose reductions. The second development is immunotherapy. The landmark KEYNOTE-522 trial established perioperative pembrolizumab, an antibody that blocks the PD-1 immune checkpoint, combined with neoadjuvant chemotherapy as the new standard of care for stage II and III disease, improving both pathologic complete response and overall survival. Yet real-world experience with that regimen has revealed higher-than-anticipated immune-related adverse events, frequent discontinuation, and dose reductions, raising concerns about how much drug burden patients can realistically tolerate.

Against this shifting backdrop, a research team led by investigators at the University of Iowa and the University of Wisconsin, working with the Greater Plains Collaborative, a PCORnet clinical research network spanning ten medical centers across nine states, asked a deceptively simple question: in actual clinical practice, does adding anthracyclines to carboplatin plus taxane chemotherapy improve outcomes? They compared two regimens among women with stage I to III triple-negative breast cancer diagnosed between 2012 and 2024. The first, abbreviated CbT, consisted of carboplatin plus a taxane. The second, AC-CbT, added doxorubicin plus cyclophosphamide, the classic anthracycline combination. The primary outcome was pathologic complete response as recorded in hospital cancer registries by treating physicians, supplemented where necessary by comparisons of pre-treatment clinical staging and post-surgical pathologic staging.

The results were striking in their clarity. Among the 878 women analyzed, 41.2 percent received the anthracycline-free CbT regimen and 58.8 percent received AC-CbT. Half of the entire cohort achieved a pathologic complete response. On the surface, the anthracycline group appeared to fare slightly better, with response rates of 51.9 percent versus 47.2 percent for CbT, but this difference was not statistically significant. More importantly, after multivariable adjustment using inverse probability weighting, accounting for age, stage, grade, race, body mass index, comorbidity, treatment site, and pembrolizumab use, the addition of anthracyclines conferred no meaningful advantage, with an adjusted odds ratio of 1.05 and a confidence interval spanning 0.69 to 1.59. Sensitivity analyses that reclassified ambiguous response data or excluded those cases entirely produced the same conclusion.

What did predict response? Pembrolizumab use nearly doubled the odds of pathologic complete response, with an adjusted odds ratio of 1.97. Higher-grade tumors were more likely to respond, paradoxically, with grade 3 disease carrying nearly twice the odds of grade 1 or 2 tumors, a pattern consistent with the principle that the most biologically aggressive cancers are often the most sensitive to chemotherapy. Lower stage predicted higher response, as expected, and women diagnosed between ages 40 and 49 were roughly twice as likely to respond as those diagnosed at 60 or older, a finding the investigators probed extensively with post hoc and machine-learning analyses without uncovering an obvious explanation. Notably, an interaction analysis showed no evidence that anthracyclines mattered more or less depending on whether pembrolizumab was given, suggesting the anthracycline question remains unresolved in both the immunotherapy era and outside it.

The study also documents a dramatic practice transformation. Among all 2,413 triple-negative breast cancer patients in the underlying dataset, use of any carboplatin-containing regimen rose from 7.1 percent in 2012 to 72.8 percent in 2024. Pembrolizumab uptake was swift following the 2021 publication of KEYNOTE-522 results, and by the end of the study period, 46 percent of the analytic cohort had received it, concentrated overwhelmingly in the anthracycline-containing group. Interestingly, the data reveal how oncologists tailor care in the real world: older patients and those with comorbidities were more likely to forgo anthracyclines, while those with stage III, more aggressive disease were more likely to receive all four cytotoxic drug classes. Nearly a third of patients received combinations that had never appeared in clinical guidelines, reflecting deliberate individualized risk-benefit balancing rather than rigid protocol adherence.

Perhaps most provocative is the finding that carboplatin has penetrated into stage I disease, a population excluded from KEYNOTE-522 and still recommended anthracycline-taxane therapy by National Comprehensive Cancer Network guidelines. Among all stage I patients in the dataset, 19.4 percent received carboplatin, rising from 3.4 percent in 2012 to 41.7 percent in 2024. Within the analytic cohort, stage I patients showed similar response rates with or without anthracyclines. This off-guideline adoption finds precedent in the adjuvant PATTERN trial, in which carboplatin plus paclitaxel matched anthracycline-based therapy in five-year overall survival, with three-quarters of enrolled patients having node-negative disease.

The investigators are careful to acknowledge limitations. As an observational study relying on electronic health records and registry data, it cannot exclude unmeasured confounding, and variables such as performance status, BRCA mutation status, treatment toxicity, adherence, and relative dose intensity were unavailable. Lower delivered doses in the four-drug regimen could theoretically explain the lack of anthracycline benefit, although the authors argue that real-world tolerability is itself a legitimate measure of a regimen’s worth. The cohort also lacked racial diversity, and a notable fraction of responses were classified as “not otherwise specified,” though sensitivity analyses confirmed the robustness of the central finding.

Even with those caveats, the study represents, to the authors’ knowledge, the largest real-world analysis of triple-negative breast cancer spanning the immunotherapy era, and it provides the strongest real-world evidence yet that anthracyclines may not earn their place alongside carboplatin-based chemoimmunotherapy. Ongoing randomized trials, most notably the SCARLET trial evaluating anthracycline-free chemoimmunotherapy against the full KEYNOTE-522 regimen, will determine whether anthracycline omission can be definitively declared noninferior. Until then, this research shifts the burden of proof: the question is no longer whether patients can safely be spared anthracyclines, but whether the drugs still have anything left to offer.

Subject of Research: Anthracycline-containing versus anthracycline-free carboplatin-based regimens in early-stage triple-negative breast cancer: real-world outcomes

Article Title: Anthracycline-containing versus anthracycline-free carboplatin-based regimens in early-stage triple-negative breast cancer: real-world outcomes

Article References: Kroll, M. R., Schroeder, M. C., Neuner, J. M., Viyyuri, B. R., Ravindra, A., McDougall, D., Chapman, C. G., Fleege, N. M. G., Chrischilles, E. A., Song, X., & Phadke, S. (2026). Anthracycline-containing versus anthracycline-free carboplatin-based regimens in early-stage triple-negative breast cancer: real-world outcomes. Breast Cancer Research and Treatment, 219(2), Article 7. https://doi.org/10.1007/s10549-026-08071-8

Image Credits: AI Generated

DOI: 10.1007/s10549-026-08071-8

Keywords: Anthracycline-containing, versus, anthracycline-free, carboplatin-based, regimens, early-stage, triple-negative, breast, cancer, real-world, outcomes, scientific research

The METABRIC cohort, used for evaluation in this study, has become one of the most widely cited resources in breast cancer computational research. It was originally assembled through the Molecular Taxonomy of Breast Cancer International Consortium, which profiled tumor specimens with gene expression microarrays and single-nucleotide polymorphism arrays while collecting detailed clinical annotations, including survival time, vital status, tumor stage, grade, and treatment information. The availability of matched clinical, transcriptomic, and copy number data for the same patients makes it unusually well suited for multimodal modeling, since many cohorts provide only one molecular layer. Its public release through cBioPortal has enabled reproducible benchmarking, and numerous prognostic modeling studies have used it, which facilitates direct comparison of new methods against previously reported performance levels.

Copy number variation, one of the three data modalities integrated by the framework, captures gains and losses of chromosomal regions that occur frequently in breast tumors and can influence prognosis by altering gene dosage. Amplification of loci such as 17q12, which contains the ERBB2 gene, and loss of regions on chromosome 17p involving TP53 are well-characterized examples with direct clinical relevance. Unlike gene expression, which reflects dynamic transcriptional activity and is sensitive to sampling and processing conditions, copy number profiles are comparatively stable measurements of genomic alteration. Including this modality alongside transcriptomic and clinical data therefore supplies the model with information about the underlying genomic architecture of the tumor, complementing the more variable expression layer and the demographic and pathological detail in the clinical record.

The reported performance gap between the hybrid framework and the conventional stacking ensemble, 0.936 versus 0.898 ROC-AUC, is meaningful in the context of prognostic modeling, where incremental gains become progressively harder to achieve as models approach the ceiling imposed by noise in the outcome labels themselves. Survival endpoints in observational cohorts are affected by treatment heterogeneity, censoring, and variation in follow-up, all of which place an upper bound on achievable predictive accuracy. Gains of this magnitude, achieved while also improving recall and reducing false negatives, suggest that the stabilization provided by the bagging branch recovers predictive signal that a single stacking pass leaves buried in prediction variance rather than merely fitting noise more aggressively.

The emphasis on reducing false-negative predictions deserves particular attention from a clinical standpoint. In prognostic stratification, a false negative means a patient at genuinely elevated risk of poor outcome is classified as low risk and may be undertreated or monitored less intensively. The consequences of missing a high-risk patient are generally considered more severe than the consequences of flagging a low-risk patient for additional surveillance, which is why sensitivity and recall are weighted heavily in clinical risk model evaluation. The observation that the hybrid framework increased sensitivity relative to the stacking ensemble on the independent test set, without sacrificing overall discrimination, indicates that the improvement is concentrated where it matters most for patient management rather than distributed evenly across easy and difficult cases.

The use of stratified tenfold cross-validation alongside an independent test set reflects a methodological practice that strengthens confidence in the reported results. Cross-validation with stratification preserves the class balance of the outcome variable in each fold, which is important because survival outcomes in breast cancer cohorts are often imbalanced, with fewer events than censored observations. Evaluating across ten folds provides an estimate of how much performance fluctuates when the training composition changes, and the reported consistency in mean ROC-AUC, F1-score, balanced accuracy, and the Matthews correlation coefficient across folds suggests that the framework’s advantage is not an artifact of a single favorable data split. The Matthews correlation coefficient is especially informative in this setting because it summarizes all four cells of the confusion matrix and remains reliable under class imbalance, unlike raw accuracy.

The challenge of integrating heterogeneous data modalities is compounded by differences in dimensionality and scale across the layers involved. Clinical variables typically number in the dozens, while gene expression panels contribute thousands of features and copy number profiles span tens of thousands of genomic loci. Naive concatenation of such matrices allows the high-dimensional molecular layers to dominate the learned representation, potentially drowning out the compact but highly informative clinical signals such as tumor stage and nodal status. Meta-learning within a stacking architecture offers a partial solution, because the meta-learner operates on base model outputs rather than raw features, effectively giving each modality a chance to be distilled into a prediction before integration and reducing the risk that one data layer overwhelms the others.

The overfitting risk inherent in high-dimensional, low-sample-size settings, often described as the large p, small n problem, is a persistent obstacle in omics research. When the number of features approaches or exceeds the number of patients, models can achieve perfect fits to training data by exploiting spurious correlations that do not generalize. This is a central reason why many published prognostic models perform well in internal validation but fail in external cohorts. The design choice of using bagging to stabilize the meta-learning stage directly targets this failure mode, since averaging over bootstrap resamples dampens the influence of any particular subset of patients or features on the final prediction, producing decision boundaries that are less sensitive to sampling idiosyncrasies.

The broader trend toward multimodal artificial intelligence in oncology, noted in the study’s positioning of its contribution, has been driven largely by imaging applications, where deep learning on histopathology slides and radiological images has attracted the most investment and clinical validation effort. Structured multi-omics prognostic modeling has progressed more quietly, in part because the data are tabular and high-dimensional rather than spatially structured, making them less amenable to the convolutional and transformer architectures that have transformed image analysis. Ensemble methods built on tree-based and classical learners remain highly competitive for tabular biomedical data, and the finding that a carefully designed ensemble of such models outperforms more elaborate alternatives reinforces the point that architectural novelty is not always the limiting factor in predictive performance.

Weighted probability fusion, the final combination step in the framework, represents a simple but principled alternative to hard voting or unweighted averaging. By assigning different weights to the stacking and bagging branches, the framework can express the relative reliability of integrated multimodal inference versus stabilized prediction, and these weights can be tuned on validation data. Soft fusion over probability outputs also preserves more information than voting on discrete class labels, since the confidence of each branch contributes to the final decision. This granularity matters at the decision thresholds used clinically, where patients near the boundary between risk strata are precisely those for whom well-calibrated probabilities, rather than binary labels, are most useful.

From the perspective of clinical translation, several considerations follow from these results. Any prognostic model intended for prospective use must demonstrate generalization beyond the cohort on which it was developed, and while the independent test set used here provides an initial check, validation in external cohorts with different patient demographics, treatment protocols, and assay platforms remains the decisive test. The reliance on gene expression and copy number data also implies that deployment requires molecular profiling infrastructure, which is increasingly routine in oncology but still unevenly distributed. The stability of performance across validation folds, together with the open availability of the underlying cohort, positions this framework as a reproducible baseline against which future multimodal prognostic methods for structured omics data can be measured, and the demonstration that complementary ensemble roles improve robustness offers a design principle likely to extend to other cancer types where matched clinical and molecular data are available.

Subject of Research: A heterogeneous multimodal ensemble framework for multi-omics breast cancer prognosis

Article Title: A heterogeneous multimodal ensemble framework for multi-omics breast cancer prognosis

Article References: Bozorgpour, R., & Sadrabadi, M. S. (2026). A heterogeneous multimodal ensemble framework for multi-omics breast cancer prognosis. Clinical Cancer Bulletin, 5(1), Article 16. https://doi.org/10.1007/s44272-026-00068-0

Image Credits: AI Generated

DOI: 10.1007/s44272-026-00068-0

Keywords: heterogeneous, multimodal, ensemble, framework, multi-omics, breast, cancer, prognosis, scientific research

Lysine demethylases, often abbreviated KDMs, are enzymes that remove methyl chemical groups from lysine residues on histone proteins, the spools around which DNA is wound inside every cell. This seemingly simple act of demethylation has profound consequences for chromatin architecture and gene expression. When KDMs malfunction, the delicate epigenetic balance that governs which genes are switched on or off can be disrupted, contributing to uncontrolled cell proliferation, evasion of programmed cell death, and the development of metastatic behavior. In breast cancer specifically, individual members of this enzyme family have previously been implicated in tumor progression and therapeutic resistance, but a holistic, multi-database portrait of the entire family has been conspicuously lacking.

To construct that portrait, the research team led by investigators at Tabriz University of Medical Sciences harnessed an impressive arsenal of publicly available bioinformatics platforms. They drew upon UALCAN for tumor versus normal expression comparisons, bc-GenExMiner for clinicopathological correlations, the Kaplan-Meier Plotter for survival analyses, cBioPortal for genomic alteration profiling, STRING and Cytoscape for protein interaction network construction, Enrichr for pathway enrichment, MethSurv for CpG methylation survival analysis, GSCALite for drug sensitivity correlations, and TIMER for immune infiltration assessments. All gene names were standardized to HGNC-approved symbols to ensure consistency across the disparate data sources, a methodological safeguard that strengthens the reliability of cross-platform integration.

The analysis revealed that KDM genes do not behave uniformly across breast cancer. Instead, they display sharply divergent expression patterns depending on the molecular subtype of the tumor. Genes including KDM1A, KDM1B, KDM4C, KDM4D, KDM4E, and KDM5A were significantly upregulated in triple-negative and basal-like breast cancers, the most aggressive and hardest-to-treat subtypes, while genes such as KDM2B, KDM3B, KDM4B, KDM6A, and KDM6B were markedly downregulated in the same contexts. Pairwise co-expression analysis within the triple-negative cohort identified two distinct modules: an oncogenic cluster anchored by a strong correlation between KDM1B and KDM5A, and a separate suppressor module linking KDM3B with PHF2. This architecture suggests that groups of demethylases may act cooperatively, either fueling or restraining tumor aggressiveness, rather than functioning as isolated molecular actors.

Survival analysis delivered some of the study’s most clinically resonant findings. Higher expression of KDM1B, KDM4A, KDM4B, KDM5A, KDM5C, KDM6A, and RSBN1 was significantly associated with improved overall survival, with KDM5C showing one of the strongest protective effects. In stark contrast, elevated KDM5B expression predicted shorter overall survival and poorer relapse-free survival, marking it as a candidate adverse prognostic marker. Relapse-free survival analysis extended this picture further, implicating elevated KDM5A, KDM5B, and KDM7A in earlier relapse while confirming the protective associations of more than a dozen other family members. Taken together, these data suggest that a patient’s KDM expression signature could one day inform risk stratification and guide the intensity of adjuvant therapy.

Genomic alteration analysis through cBioPortal, drawing on more than one thousand invasive breast carcinoma samples, identified KDM5B as the most frequently altered KDM gene, with amplification emerging as the dominant mechanism of dysregulation in roughly eighteen percent of tumors. Crucially, alterations in KDM5B, KDM2A, and KDM4C were strongly enriched in metastatic disease, aggressive histologic subtypes including infiltrating ductal and metaplastic carcinomas, and hormone receptor-negative phenotypes. Many of these amplifications and missense mutations clustered within functional domains such as the JmjC catalytic core and DNA-binding regions, implying that the alterations are not passenger events but likely enhance demethylase activity in ways that actively promote tumor progression. Protein interaction network modeling pinpointed KDM6A, KDM6B, KDM1A, KDM1B, and PHF8 as the central regulatory hubs within the broader epigenetic landscape of breast cancer.

Perhaps the most therapeutically provocative dimension of the study involves drug sensitivity. Cross-referencing KDM expression with large pharmacogenomic datasets from the Genomics of Drug Sensitivity in Cancer and Cancer Therapeutics Response Portal revealed that most KDM genes correlate negatively with drug sensitivity, meaning their overexpression may actively foster chemotherapy resistance. KDM2B stood out as the linchpin, showing the strongest and most consistent correlations across both datasets. Elevated KDM2B expression was linked to reduced sensitivity to Methotrexate, Doxorubicin, PX-12, I-BET-762, and Belinostat, clinically relevant agents spanning antifolate, anthracycline, redox-targeting, bromodomain-inhibiting, and HDAC-inhibiting classes. Prior experimental work in glioblastoma and other models has shown that KDM2B inhibition impairs cancer stem-like cell survival by inducing DNA damage and apoptosis, providing a plausible mechanistic rationale for how targeting this enzyme could resensitize tumors to standard chemotherapy.

The study also uncovered a striking methylation dimension. Using MethSurv, the researchers identified dozens of individual CpG sites within KDM genes whose methylation status was significantly associated with patient survival, including thirty-five prognostically relevant CpG sites in KDM2B alone and forty-six in KDM4B. This epigenetic crosstalk between histone demethylase genes and their own DNA methylation patterns underscores the layered complexity of cancer epigenetics and hints at potential biomarker applications beyond simple mRNA measurement. Meanwhile, enrichment analysis confirmed that KDM family genes are centrally involved in histone lysine demethylation, chromatin remodeling, 2-oxoglutarate-dependent dioxygenase activity, and metabolic pathways, connecting these enzymes to the metabolic reprogramming that is increasingly recognized as a hallmark of malignancy.

Immune infiltration analysis through TIMER added yet another layer of significance. KDM5A, KDM6A, and KDM7A expression correlated strongly and positively with CD8-positive T cell infiltration, suggesting these demethylases may help shape anti-tumor immunity. KDM2B showed notable associations with dendritic cells, neutrophils, CD4-positive T cells, and B cells, while KDM6A correlated with macrophage presence. Given the central role of tumor immune microenvironment composition in determining response to immunotherapy, these correlations suggest that KDM expression profiles could eventually help predict which patients are most likely to benefit from immune checkpoint inhibitors. The authors are careful to acknowledge the principal limitation of their work: it is entirely computational, relying on public databases without wet-lab experimental validation. Future studies employing in vitro knockdown assays and in vivo models will be essential to confirm these associations and dissect the underlying molecular mechanisms. Nevertheless, by synthesizing expression, survival, mutation, methylation, drug sensitivity, and immune data into a single unified framework, this investigation delivers the most complete picture to date of how the lysine demethylase family operates in breast cancer, and it does so with a clarity that should accelerate the translation of epigenetic insights into genuinely personalized therapeutic strategies.

The biological significance of lysine demethylation stems from the fact that methylated lysine residues on histone tails can carry distinct methylation states — mono-, di-, or trimethyl — and each state conveys a different transcriptional message. The KDM family reflects this complexity, comprising two mechanistically distinct classes: the flavin adenine dinucleotide-dependent amine oxidases such as KDM1A and KDM1B, and the larger group of JmjC-domain proteins that require iron and 2-oxoglutarate as cofactors. This enzymatic split has practical consequences for drug discovery, since the two classes are inhibited through entirely different chemical strategies, and several small-molecule demethylase inhibitors are already being evaluated in early-phase clinical trials for hematologic and solid malignancies.

The subtype-specific expression patterns observed in this study align with a broader principle in breast cancer biology: epigenetic regulators frequently act in a context-dependent fashion, behaving as oncogenes in one molecular background and tumor suppressors in another. This duality helps explain why broad demethylase inhibition may not be a viable therapeutic strategy, and why a precision approach — one that profiles a patient’s tumor for specific KDM alterations before selecting an epigenetic drug — is likely to be more fruitful.

The strong correlations between KDM expression and immune cell infiltration also connect the family to the emerging field of epigenetic immunomodulation. Chromatin regulators are increasingly understood to control the expression of immune checkpoints, antigen presentation machinery, and cytokine signaling pathways, and combining epigenetic drugs with immune checkpoint blockade has become an active area of clinical investigation. The observed links between KDM5A, KDM6A, and CD8-positive T cell infiltration provide a computational foundation for testing such combinations in breast cancer settings.

It is worth noting that the study’s findings derive entirely from curated patient cohorts within platforms built on The Cancer Genome Atlas and related resources, and the reported effect sizes — such as the hazard ratio of 1.5 for KDM5B — represent statistical associations rather than demonstrated causation. Translating these computationally derived hypotheses into clinical utility will require laboratory validation and, ultimately, prospective testing of KDM-based biomarkers in independent patient populations.

Subject of Research: Integrated bioinformatics profiling of the lysine demethylase gene family in breast cancer

Article Title: Integrated bioinformatics profiling of the lysine demethylase gene family in breast cancer

Article References: Mirzaei, Z., Barati, T., Ebrahimi, A., & Khaniani, M. S. (2026). Integrated bioinformatics profiling of the lysine demethylase gene family in breast cancer. Discover Biotechnology, 3(1), Article 9. https://doi.org/10.1007/s44340-026-00055-0

Image Credits: AI Generated

DOI: 10.1007/s44340-026-00055-0

Keywords: Integrated, bioinformatics, profiling, lysine, demethylase, gene, family, breast, cancer, scientific research