Invasive Breast Carcinoma

This study details the analytical validation of the Illumina TruSight Oncology 500 (TSO500) assay within a CAP-certified clinical laboratory in Brazil. It also presents real-world findings from Comprehensive Genomic Profiling (CGP) performed on 454 patients with solid tumors. The research highlighted the genomic landscape of these tumors, including the detection of both novel and canonical gene fusions. The analytical performance of the assay was assessed, demonstrating high specificity and positive predictive value. The authors conclude that TSO500 provides crucial insights for diagnosis, prognosis, and treatment decisions in oncology.

This study investigates predictive biomarkers for immune checkpoint inhibitor (ICI) response in metastatic urothelial carcinoma. Researchers found that ICI responders exhibited higher tumor mutational burden (TMB) and enriched mutations in genes such as PIK3CA. Functional in vitro and in vivo studies demonstrated that PIK3CA mutations enhance tumor immunogenicity by activating the IRF1-NLRC5-MHC-I axis, thereby improving antigen presentation and CD8+ T-cell cytotoxic response. These findings suggest that PIK3CA mutation could serve as a biomarker to predict ICI sensitivity and represents a novel immune-modulating mechanism.

This study introduces MOCDT, an innovative bioinformatics framework for multi-cancer detection and tissue-of-origin classification using cell-free DNA (cfDNA). MOCDT employs a two-stage pipeline: high-specificity cancer detection followed by conditional tissue-of-origin classification. The model integrates a supervised multi-modal autoencoder with a patient similarity network and a Graph Convolutional Network for relational learning. Applied to a cohort of healthy controls and eight cancer types, MOCDT achieved 95.74% specificity and 96.22% sensitivity for cancer detection, alongside 75.2% Top1 and 91.06% Top3 accuracy for tissue-of-origin classification.

Advances presented at AACR 2026 highlight a shift towards integrated, agentic artificial intelligence (AI) systems in oncology. Platforms like Synapse facilitate large-scale data coordination, while conversational and multi-agent tools enable simplified interaction with multimodal cancer data. AI demonstrates high accuracy and scalability for real-world data transformation, including clinical document abstraction and social determinants of health analysis. Clinically, it enhances imaging biomarkers, trial matching, and cohort identification, while also accelerating therapeutic discovery, including CAR-T development and immunotherapy target identification.

This study comprehensively compared seven homologous recombination deficiency (HRD) prediction methods in 235 patients with early-stage triple-negative breast cancer. The methods included sequencing-based, copy number-based, functional, mRNA-based, and image-based approaches, with an exploratory comparison to the FDA-approved Myriad myChoice CDx assay. Results revealed substantial concordance among methods, alongside specific discordances attributed to data preprocessing and training strategies. Despite these variations, all methods demonstrated similar prognostic performance in patients treated with adjuvant chemotherapy. The study emphasizes the critical need for rigorous optimization of data processing workflows and threshold definitions to ensure consistency and reproducibility of HRD classifications.

Accurate identification of somatic variants is crucial in oncology, yet existing methods often struggle with high sensitivity in certain genomic regions. VariantMedium, a novel tool, combines a tree-based classifier with a 3D DenseNet convolutional neural network. This model was trained and validated on a large dataset of experimentally confirmed variants and further refined using an active learning strategy. It demonstrates superior sensitivity and comparable or better F1 scores than benchmark tools, especially in genomic regions prone to high sequencing error rates.

This study assesses the performance of in silico prediction tools for genetic variant curation within a panel of cancer predisposition genes. Researchers applied ClinGen SVI Working Group recommended thresholds and AlphaMissense predictions to variants in genes such as BRCA1, BRCA2, TP53, TERT, and ATM, which have established pathogenicity or benignity. The findings indicated insufficient sensitivity for pathogenic TERT variants and benign TP53 variants. The study emphasizes that the performance of these tools can be gene-specific and is influenced by their training datasets. Consequently, it is crucial to validate these tools for individual genes, especially for missense variants.

Accurate RNA splicing is vital for gene expression, yet its mechanisms and the impact of mutations, which can cause diseases like cancer, are not fully understood. Current prediction tools struggle with long-range dependencies and often lack interpretability. SpliceSelectNet (SSNet) is a novel hierarchical Transformer-based deep learning model designed to predict splice sites from DNA sequences up to 100 kb. It integrates local and global attention mechanisms to efficiently capture both proximal and distal regulatory signals. SSNet achieves state-of-the-art performance in splice site prediction and aberrant splicing detection, offering a biologically interpretable framework for modeling long-range splicing regulation.

This study investigated the impact of next-generation sequencing (NGS) panel size on therapeutic recommendations in precision oncology. Comparing a 430-gene panel to a smaller 185-gene panel in 281 patients, researchers found molecular alterations in nearly all tumors. While most stratified therapy recommendations were achievable with the smaller panel, the expanded panel enabled additional recommendations for 8.8% of patients and identified actionable variants in five extra genes. The study emphasizes the necessity of sufficient genomic coverage for reliable tumor mutational burden (TMB) calculation, thereby establishing a minimal standard for genomic cancer care.

The Oncogenicity Variant Interpreter (OncoVI) is an open-source, Python-based tool developed to harmonize and automate the oncogenicity classification of somatic variants in precision oncology. It implements guidelines from the Clinical Genome Resource/Cancer Genomics Consortium/Variant Interpretation for Cancer Consortium, performing functional annotation and evidence collection from public resources. OncoVI achieved 80% accuracy on a gold standard set of 93 somatic variants and showed 79% concordance with Molecular Tumor Board assessments on 7802 real-world variants. This tool aims to support reproducible and standardized somatic variant interpretation across institutions.

This study performed whole genome and transcriptome sequencing on 50 tumor samples from 37 patients with locally advanced or metastatic breast cancer. It uncovered extensive genomic complexity, with triple-negative breast cancer (TNBC) showing the highest tumor mutational burden. Homologous recombination deficiency (HRD) was found in 27% of patients, frequently in BRCA1/2 wild-type cases due to deleterious structural variants in other repair genes. Therapeutically actionable alterations were identified in 84% of patients, including a novel ESR1::EP300 fusion potentially linked to endocrine resistance. These findings underscore the utility of whole-genome sequencing for characterizing metastatic disease and guiding therapies.

Somatic mutations in GC-rich promoter regions are significant drivers of cancer, yet their detection is challenging due to poor sequencing coverage in these areas. This study introduces a hybrid capture assay optimized for over 3000 cancer gene promoters, enabling deep sequencing of these complex regions. This method facilitates the discovery of reliable point mutations, short insertions/deletions, copy number variants, and mutational signatures. The assay nominated candidate noncoding driver mutations in CDK4, SMAD3, and GATA3 in breast cancer, paving the way for future functional follow-up.

This review examines established and emerging biomarkers for stratifying patients with early-stage breast cancer to optimize adjuvant systemic therapy. It highlights that while clinicopathological factors remain fundamental, decision-making is increasingly driven by precise biological markers. ER and HER2 status are crucial, and multigene assays refine recurrence risk and chemotherapy benefit in hormone receptor-positive cancers. Immune and DNA-repair biomarkers inform targeted therapies in HER2-positive and triple-negative subtypes, while mutation profiling of genes like ESR1, PIK3CA, AKT, MTOR, and PTEN guides targeted treatments. Emerging approaches, including liquid biopsy and artificial intelligence, offer dynamic insights but require prospective validation.

This meta-analysis assessed the performance of breast cancer risk prediction models in women with a family history of the disease. The Gail (BCRAT) and BOADICEA models demonstrated good calibration, while Tyrer-Cuzick overpredicted and BRCAPRO underpredicted risk. Regarding discriminatory accuracy, no single model was clearly superior, with Tyrer-Cuzick version 8, BOADICEA, and BRCAPRO showing similar modest discrimination. The authors suggest that BOADICEA may be useful for patient management in familial breast cancer risk settings, despite limitations such as high risk of bias and study heterogeneity.

This prospective study investigated the prevalence of non-BRCA germline pathogenic variants (PVs) and their impact on response to neoadjuvant therapy (NAT) in patients with triple-negative breast cancer (TNBC). Among 184 patients, 36% harbored PVs in pan-cancer susceptibility genes, including variants in 31 genes not previously found in other US cohorts. Patients received NAT with doxorubicin, cyclophosphamide, paclitaxel, and carboplatin, with or without immunotherapy or targeted therapy. Despite this high prevalence, the presence of these non-BRCA germline PVs was not associated with a significant difference in pathologic complete response (pCR) or radiologic response. The findings suggest the utility of multigene panels for screening but indicate that these variants are not predictive of NAT response in TNBC.

Synthetic lethality describes a genetic interaction where two specific genetic alterations together impair cell viability, while either alteration alone is compatible with survival. This concept offers promising therapeutic avenues for targeting previously undruggable cancer pathways. High-throughput screening technologies have facilitated the discovery of novel druggable synthetic lethal vulnerabilities, particularly in DNA damage response and epigenetic alterations. Clinically approved PARP inhibitors have validated this approach in BRCA-mutant cancers. Emerging strategies targeting PRMT5, MAT2A, WRN, PKMYT1, and WEE1 are currently undergoing late preclinical or early clinical evaluation.

This article discusses the increasing complexity of therapeutic decision-making for early-stage, hormone receptor-positive, HER2-negative breast cancer. It highlights the significance of advances in tumor biology, particularly genomic assays, in improving risk stratification and identifying patients most likely to benefit from chemotherapy. The review also explores current challenges and controversies, such as the use of genomic assays in premenopausal women with node-positive disease. Furthermore, it examines the emerging role of circulating tumor DNA as a prognostic and predictive biomarker.

Data from the 2025 San Antonio Breast Cancer Symposium highlight the expanding role of circulating tumor DNA (ctDNA) in breast cancer management. ctDNA is now being utilized to guide therapeutic decisions, particularly in advanced hormone receptor-positive breast cancer. The SERENA-6 trial demonstrated that ctDNA-detected ESR1 mutations enabled adaptation of endocrine therapy, leading to improved clinical outcomes. Furthermore, in early-stage disease, ctDNA identifies minimal residual disease (MRD) and patients at high risk of recurrence. These advancements position ctDNA as a crucial biomarker for precision oncology and dynamic treatment adaptation.

Improved survival rates for reproductive-age breast cancer patients highlight significant oncofertility challenges. Fertility preservation, gonadotoxicity, and pregnancy safety are key concerns requiring early and integrated management within the therapeutic plan. Prompt referral for fertility preservation is crucial, ensuring it does not compromise oncologic outcomes. Modern therapies such as PARP inhibitors, CDK4/6 inhibitors, HER2-monoclonal antibodies, and endocrine treatments raise specific questions due to a lack of prospective human fertility data. A precautionary approach is recommended, including effective contraception and adherence to washout periods.

This meta-analysis investigated the efficacy of rechallenging with CDK4/6 inhibitors (CDK4/6i) after progression in patients with HR-positive, HER2-negative advanced/metastatic breast cancer. The study included 1396 patients from eight clinical trials, comparing CDK4/6i rechallenge plus endocrine therapy to endocrine therapy alone. Results indicated a significant improvement in median progression-free survival (PFS) of 5.8 months versus 3.7 months. A greater benefit was observed when switching to a different CDK4/6 inhibitor, particularly with abemaciclib, but no overall survival benefit was demonstrated.

The treatment paradigm for hormone receptor-positive (HR+), HER2-negative (HER2-) metastatic breast cancer (MBC) has significantly evolved from sequential therapies to more personalized strategies. CDK4/6 inhibitors combined with endocrine therapy remain the first-line standard of care, consistently improving progression-free survival and, for some agents, overall survival. The emergence of antibody-drug conjugates (ADCs) such as trastuzumab deruxtecan, sacituzumab govitecan, and datopotamab deruxtecan is reshaping the therapeutic landscape, including for HER2-low and HER2-ultralow disease. Metastasis-directed therapy is also considered for oligometastatic or oligoprogressive disease, requiring nuanced decision-making. Overall management now integrates tumor biology, therapeutic advances, and patient-defined goals of care.

This nationwide Japanese study resolved the status of the BRCA2 c.7847C>T (p.Ser2616Phe) variant, which is specific to the Japanese population and was previously of uncertain significance. By integrating quantitative cosegregation analyses, robust functional evidence, and population frequency data, researchers accumulated strong evidence. The evaluation led to the reclassification of this variant as "Pathogenic" according to the ClinGen ENIGMA framework. This reclassification is crucial for ensuring eligible Japanese patients gain access to appropriate targeted therapies, such as PARP inhibitors.

This paper introduces VBVarSel, a novel annealed variational Bayes algorithm designed for high-dimensional data analysis. It aims to overcome challenges associated with irrelevant variables and the selection of an appropriate number of clusters. The algorithm is particularly suited for disease subtyping and biomarker discovery in omics datasets. The authors demonstrate that VBVarSel outperforms existing methods in both simulated and real biomedical examples, specifically for cancer subtyping.

Liquid biopsy represents a significant advancement in oncology, providing a minimally invasive alternative to tissue biopsy for cancer diagnosis and patient monitoring. It utilizes biomarkers such as circulating tumor DNA (ctDNA) for early cancer detection, minimal residual disease (MRD) surveillance, and dynamic treatment response monitoring. Clinical trials have demonstrated its effectiveness in earlier recurrence detection and more precise guidance for therapeutic decisions. Despite challenges related to sensitivity and technical variability, the integration of artificial intelligence and multi-omic approaches promises to enhance its future performance.

This study compares the performance of large language models (LLMs such as GPT-4.1-mini and Gemini-2.5-Flash) and machine learning (ML) algorithms (decision tree and XGBoost) for classifying biomedical literature according to the CIViC evidence level system. LLMs were evaluated using zero- and few-shot prompting strategies, while ML models utilized TF-IDF and word embedding representations. XGBoost with TF-IDF achieved the highest performance (micro-F1 = 0.83), outperforming both LLMs and decision trees. All models performed best on mid-range evidence levels (B to D) but struggled with high (A) and inferential (E) levels. The findings suggest that abstract-level evidence classification is largely driven by explicit lexical cues, with limited additional benefit from standalone LLM-based approaches.