Breast Cancer Now

Methods to detect circulating tumor DNA (ctDNA) enable minimally invasive responsive monitoring of cancer dynamics. However, sensitive and cost-effective methods are still lacking. Current methods for detecting cancer signals in shallow whole-genome sequencing (sWGS) data from cell-free DNA (cfDNA) via copy number aberration (CNA) analysis typically have a limit of detection of approximately 3% tumor fraction (TF). We developed informCNA, a bioinformatics method that leverages CNA information from sWGS of tumor or pre-treatment plasma samples with high TF as references, enabling ctDNA detection down to 0.2% TF across multiple cancer types. In 177 serial plasma samples from 18 patients with ovarian cancer, informCNA showed high concordance with the standard serum protein marker CA-125 and identified recurrence a median of 3.7 months earlier than CA-125 test. These results demonstrate the potential of personalized CNA analysis through sWGS for estimating ctDNA burden, enabling precise and cost-effective disease monitoring and early detection of relapse.

Abstract Background: Detection of circulating tumour DNA (ctDNA) in plasma predicts future recurrence in early breast cancer (EBC). Previous studies have relied on tumour-informed approaches with personalised assays. In this study, we performed orthogonal testing on plasma samples collected prospectively in the c-TRAK-TN study to compare a tissue-free assay with digital droplet PCR (ddPCR) for ctDNA analysis. Methods: c-TRAK-TN recruited 161 patients with early triple negative breast cancer (TNBC) deemed clinically high risk of relapse. Patients underwent ctDNA analysis by ddPCR starting within 4 weeks of completion of treatment and continuing every 3 months for up to 2 years. ddPCR tracked 1 or 2 patient specific variants as previously described. We further analysed 1062 timepoints from a total of 159 patients (median 7 per patient, range 1-11) using Guardant Reveal assay, a tissue-free approach that exploits differential methylation patterns in breast cancer compared with normal tissue. Results: Samples were successfully analysed from 95.3% (1012/1062) timepoints from 159 patients with the tissue-free Reveal assay. CtDNA was detected in 34.0% (54/159) of patients in at least 1 timepoint, with a median methylation tumour fraction of 0.266% (range 0.002 to 51.0%). Detection of ctDNA with the tumour agnostic assay strongly associated with future risk of relapse (HR 20.62, p<0.0001; logrank). Of patients who relapsed by 24 months follow up, 89.5% (34/38) of relapses were detected by prior ctDNA analysis. There were 159 patients available for comparison between ddPCR and the tissue-free assay (Reveal), in 1005 timepoints (median 7 per patient, range 1-11). Concordant test results were observed in 950/1005 timepoints with an overall agreement rate of 94.5%. In this cohort, 63.5% (101/159) of patients did not have ctDNA detected by either assay; 26.4% (42/159) had ctDNA detected by both assays and 10.1% (16/159) had discordant results. Of those detected, 44.8% (26/58) were detected by the tissue-free assay first and 6.9% (4/58) by ddPCR (p<0.0001, Fisher exact test). The median time from first detection of ctDNA to relapse for the tissue-free assay was 7.9 months (95% confidence interval (CI): 5.7-10.5 months) compared to ddPCR assay with 5.7 months (95% CI: 2.9-9.7 months) (HR = 0.6, p=0.0574; mixed effects cox model). Conclusion: Detection of ctDNA with a tissue-free assay anticipated relapse with high accuracy, in patients with early TNBC. The tissue-free assay frequently detected ctDNA at an earlier timepoint than ddPCR, trending towards a longer lead time from detection of ctDNA to relapse. Tissue-free approaches might offer simpler workflows for ctDNA analysis than tumour-informed approaches, and further assessment in clinical trials is warranted. Citation Format: N. Cunningham, R. J. Cutts, C. Swift, K. Dunne, M. Dewan, L. Kilburn, K. Goddard, P. Hall, C. Harper-Wynne, T. Hickish, I. Macpherson, A. Okines, A. Wardley, S. Waters, C. Palmieri, M. Winter, J. Bliss, D. Dustin, M. Ellis, I. Garcia- Murillas, N. C. Turner. Tissue-free circulating tumour DNA detection in patients with early triple negative breast cancer from the c-TRAK-TN trial [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr RF3-05.

Relapse and treatment resistance are common in children with high-risk neuroblastoma, and novel therapies are needed. Conventional drug discovery is slow, expensive, often fails in practice, and consequently falls short in addressing pediatric and rare conditions. In such instances, drug repurposing is a promising strategy. Here, we used two independent in silico prediction tools including machine learning to identify approved drugs for repurposing against neuroblastoma. The combination of statins and phenothiazines showed strong synergistic effects in human neuroblastoma organoids, decreased tumor growth, and prolonged survival in MYCN-amplified neuroblastoma patient-derived xenografts. The drug combination altered cholesterol metabolism through two different mechanisms and induced a phenotypic change toward an adrenergic state in vitro, which was associated with enhanced chemosensitivity. Integration of the drug combination into standard-of-care chemotherapy regressed tumors and prolonged survival in chemoresistant patient-derived xenografts. Thus, a combination of safe and approved medications added to standard-of-care chemotherapy outperforms chemotherapy alone in chemoresistant neuroblastoma.

The immune system is hypothesized to contribute to the onset of endometriosis lesions. However, the precise mechanisms underlying its role are not yet known. We introduce a novel compartmental model that describes the interactions between innate immune cells, specifically macrophages and natural killer cells, and endometrial cells, occurring within the peritoneal fluid during the early stages of (superficial peritoneal) endometriosis lesion onset. Our study focuses on retrograde influx, immune detection and immune clearance. Results show an increased influx of endometrial cells into peritoneal fluid correlates with heightened pro-inflammatory macrophage activation, but does not lead to an increase in disease. We compare the system’s response to changes in immune cytotoxicity and ability to detect ectopic endometrial cells. We predict that reduced cytotoxicity is a key driver of disease. These findings align with the increased immune activation observed clinically. Finally, we predict that an individual can transition to a diseased state following a reduction in immune system cytotoxicity and/or reduced ability to detect ectopic cells. Due to hysteresis, a significant improvement is then required to restore an individual to the disease-free state. This work provides a valuable framework to explore hypotheses of endometriosis lesion onset and assist in understanding of the disease.

Abstract Machine learning models in cancer medicine are powerful but can suffer from a lack of robustness and interpretability that limits their medical and scientific utility. Here, we introduce a new class of models for high-dimensional molecular data that is guided by interventional auxiliary information to learn latent representations that are informative, interpretable-by-design and, as we show, allow for strong generalization to new data distributions and settings. Specifically, the proposed models use causal information from genetic loss-of-function screens to guide learning and thereby discover representations that are generally useful, even far beyond the original training context. Empirical results using cancer cell line datasets support the hypothesis that causal guidance permits strong generalization and we find that the models can be applied in an entirely “zero-shot” fashion—with no fine-tuning or adaptation whatsoever—to entire cancer types that were not encountered during training. Using several large patient cohorts, we show that this strong generalization applies also in the “bench-to-bedside” context of transferring cell line-based models to clinical data. Here, we find that our models, with all representation learning steps trained only on cell line data, can be efficiently transferred to patient data, hence opening up new ways to drive clinical modelling using tractable laboratory assays. Our results provide a new and very general way of leveraging causal and interventional information to construct data-efficient and robust AI-based predictive models for molecular medicine.

[This corrects the article DOI: 10.3389/fvets.2020.00439.].

Clonal fitness and plasticity drive cancer heterogeneity. We used expressed lentiviral-based cellular barcodes combined with single-cell RNA sequencing to associate single-cell profiles with in vivo clonal growth. This generated a significant resource of growth measurements from over 20,000 single-cell-derived clones in 110 xenografts from 26 patient-derived breast cancer xenograft models. 167,375 single-cell RNA profiles were obtained from 5 models and revealed that rare propagating clones display a highly conserved model-specific differentiation program with reproducible regeneration of the entire transcriptomic landscape of the original xenograft. In 2 models of basal breast cancer, propagating clones demonstrated remarkable transcriptional plasticity at single-cell resolution. Dichotomous cell populations with different clonal growth properties, signaling pathways, and metabolic programs were characterized. By directly linking clonal growth with single-cell transcriptomes, these findings provide a profound understanding of clonal fitness and plasticity with implications for cancer biology and therapy.

Genomic defects caused by truncating mutations or deletions in the Retinoblastoma tumor suppressor gene (RB1) are frequently observed in many cancer types leading to dysregulation of the RB pathway. Here, we propose an integrative proteogenomic approach that predicts cancers with dysregulation in the RB pathway. A subset of these cancers, which we term as "RBness," lack RB1 genomic defects and yet phenocopy the transcriptional profile of RB1-defective cancers. We report RBness as a pan-cancer phenomenon, associated with patient outcome and chemotherapy response in multiple cancer types, and predictive of CDK4/6 inhibitor response in estrogen-positive breast cancer. Using RNA interference and a CRISPR-Cas9 screen in isogenic models, we find that RBness cancers also phenocopy synthetic lethal vulnerabilities of cells with RB1 genomic defects. In summary, our findings suggest that dysregulation of the RB pathway in cancers lacking RB1 genomic defects provides a molecular rationale for how these cancers could be treated.