Breast cancer is a molecularly heterogeneous disease. The authors aimed to interrogate whether transcriptional programs associated with histone deacetylases (HDACs) carry clinically relevant prognostic information. The report frames HDAC-associated molecular heterogeneity as incompletely understood with respect to downstream prognostic significance and seeks to construct and validate a transcriptional prognostic signature linked to HDAC biology.
The authors performed unsupervised and supervised analyses on transcriptomic data to derive an HDAC-associated prognostic model. Key analytic steps reported were:
Application of non-negative matrix factorization (NMF) to the TCGA-BRCA cohort to identify HDAC-related molecular subtypes and consensus clusters.
Characterization of cluster-specific features, including survival outcomes and patterns of immune infiltration, in the TCGA-BRCA dataset.
Use of weighted gene co-expression network analysis (WGCNA) to identify subtype-associated co-expression modules.
Differential-expression analysis to obtain HDAC-cluster-associated differentially expressed genes (DEGs).
Intersection of DEGs with WGCNA-derived hub genes to generate a list of candidate features for model building.
Construction of a prognostic signature using multiple machine-learning algorithms; the final 24-gene model was based on a combination identified as CoxBoost + SuperPC.
Validation of the resulting HDAC-related risk score (HRS) across the TCGA training cohort and seven external breast cancer validation cohorts.
Additional statistical assessments including multivariable and continuous-risk Cox regression to evaluate the independent prognostic value of HRS, and correlation analyses between HRS and expression of HDAC family members.
Single-cell gene-set scoring was applied using Seurat’s AddModuleScore function to evaluate cell-type–specific heterogeneity of the model-related signature score.
Preliminary experimental validation of selected hub genes was reported using quantitative real-time polymerase chain reaction (qRT-PCR) and western blotting (WB).
Note: The source abstract does not report specific sample sizes, parameter settings for NMF/WGCNA, or the identities of all evaluated external cohorts; those details were not reported in the abstract.
The study identified five HDAC-related molecular clusters in the TCGA-BRCA cohort by NMF consensus clustering. These clusters exhibited distinct overall survival outcomes and differences in immune-infiltration patterns as measured in the TCGA dataset.
WGCNA detected co-expression modules associated with the HDAC-defined subtypes. By intersecting the subtype-associated DEGs with WGCNA hub genes, the authors derived candidate features for prognostic modeling.
Using a machine-learning pipeline, a 24-gene HDAC-related risk score (HRS) model (CoxBoost + SuperPC) was constructed. The model showed stable prognostic performance in the TCGA training cohort and retained predictive validity across seven external breast cancer validation cohorts according to the abstract.
Multivariable Cox regression analyses in several major cohorts supported the HRS as an independent prognostic indicator after adjustment for available covariates (details of covariates and cohort-by-cohort statistics were not provided in the abstract).
Correlation analyses examined relationships between HRS and expression of individual HDAC family members, indicating an evaluative link between the risk score and HDAC expression patterns (specific correlation coefficients and HDAC members assessed are not reported in the abstract).
Single-cell transcriptomic analysis using Seurat’s AddModuleScore revealed that the HRS-related signature score was heterogeneous across cell types, demonstrating cell-type–specific variation in signature expression at single-cell resolution.
Among the hub genes examined, four — UTRN, LIMCH1, ARNT2, and PYDC1 — exhibited concordant expression patterns in public datasets and in the authors’ preliminary experimental assays. The report indicates these four genes were further assessed by qRT-PCR and western blotting, and that those assays showed concordant expression with the public data; the abstract does not provide quantitative experimental results or methodological details for those assays.
The authors conclude that the HDAC-associated 24-gene HRS model may serve as an exploratory prognostic framework for breast cancer. They emphasize that while the model demonstrated reproducible prognostic performance across multiple cohorts and shows cell-type–specific heterogeneity at single-cell resolution, further mechanistic investigation and prospective validation are required before translation into clinical practice.
The study positions HDAC-associated transcriptional programs as a potential source of prognostic biomarkers in breast cancer and reports initial cross-cohort validation and preliminary experimental support for selected hub genes.
The authors declared no known competing financial interests or personal relationships that could have influenced the reported work.