Identification and validation of a refined CAF-Associated diagnostic signature in breast cancer

Sep 22, 2026

Upload by Patrick

Breast cancer remains a major global health challenge, with high incidence and mortality among women. The tumor microenvironment, especially cancer-associated fibroblasts (CAFs), is increasingly recognized as a key driver of tumor progression, yet CAF heterogeneity and its diagnostic implications remain insufficiently understood. This study applied machine learning to identify CAF-associated feature genes and build a high-precision diagnostic model for breast cancer. Single-cell analyses characterized CAF heterogeneity and predicted drug sensitivity across CAF subsets, while immunohistochemical validation supported the expression patterns of refined biomarkers (FXYD1, SULF1, TNXB).

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages18

Reading Time45 min

AccessFree to download

Introduction

Tumor microenvironment and CAFs in breast cancer

Need for refined markers and diagnosis

Methods and Modeling

Machine learning feature gene identification

High-precision diagnostic model

Single-cell analysis for CAF heterogeneity and drug sensitivity prediction

Immunohistochemical validation

[www. nature.com/scientificreports](www. nature.com/scientificreports) OPEN Identification and validation of a refined CAF-Associated diagnostic signature in breast cancer Xin Zhou1, Na Wang2, Ling Shi2, Dongxin Wei1, Xiaoqin Sun2, Mingxiu Shao3, Liang Tian1, Xiaolong Guo1, Fangyuan Zhang1 & Hui Lyu3􀀍 Breast cancer remains a major global health challenge with high incidence and mortality rates among women. Recent studies have highlighted the critical role of the tumor microenvironment, particularly cancer-associated fibroblasts (CAFs), in tumor progression. However, current understanding of CAFs heterogeneity and its implications for breast cancer diagnosis and treatment remains limited. This study aimed to identify and validate refined marker genes for CAFs and to develop a diagnostic model to improve breast cancer diagnosis and therapeutic strategies. We employed various machine learning algorithms to identify feature genes associated with CAFs. Based on these genes, we constructed a high-precision diagnostic model for breast cancer. Furthermore, through single-cell analysis, we delved into the heterogeneity of CAFs and predicted the sensitivity of different CAF subsets to specific drugs. To validate the expression of these characteristic genes, immunohistochemical (IHC) experiments were also conducted. This study used machine learning to identify FXYD1, SULF1, and TNXB as refined biomarkers for CAFs in breast cancer. Among these evaluated algorithms, the Random Forest algorithm distinctly stood out as the best due to its robust classification accuracy and stability. Single-cell analysis provided insights into the heterogeneity of CAFs between Luminal and non-Luminal breast cancer, thereby enhancing our understanding of the tumor microenvironment. Drug sensitivity predictions indicated that distinct CAF subsets responded differently to specific drugs, laying a solid foundation for the development of personalized breast cancer treatment strategies. Through IHC, the expression patterns of these three biomarkers were verified: FXYD1 was expressed in myoepithelial and fibroblasts in normal breast tissue but was significantly absent in breast cancer; SULF1 was upregulated in fibroblasts of breast cancer; while the expression of TNXB did not exhibit notable variations between normal and cancerous tissues. These findings not only highlight the crucial roles played by FXYD1, SULF1, and TNXB in the development of breast cancer, but also uncover the heterogeneity CAFs. Consequently, our research provides a fresh perspective and a solid theoretical basis for advancing both early and precise diagnostic methods, as well as tailored therapeutic strategies. Keywords Cancer-associated fibroblasts, Breast cancer, Machine learning, Immunohistochemistry, Diagnostic model Abbreviations α-SMA Alpha Smooth Muscle Actin BC Breast Cancer CAF Cancer-Associated Fibroblasts CNA Copy Number Alteration DAB 3,3’-Diaminobenzidine DEGs Differentially Expressed Genes EDTA Ethylenediaminetetraacetic Acid SSC sensitivity signature collection ER Estrogen Receptor FXYD1 fxyd domain-containing transport regulator 1 1Department of Breast and Thyroid Surgery, Zibo Maternal and Child Health Hospital, Zibo, Shandong, China. 2Department of Pathology, Zibo Maternal and Child Health Hospital, Zibo, Shandong, China. 3Clinical Laboratory, Zibo Maternal and Child Health Hospital, Zibo, Shandong, China. 􀀍 email: [Lyuhui080806@163.com](Lyuhui080806@163.com) [www. nature.com/scientificreports/](www. nature.com/scientificreports/) GAM GEO GLM GO HER2 HSPG IHC iCAFs KEGG KNN LDA LR mCAFs METABRIC NB OS PCA PR ROC RFE RF RFSSULF1 SULF2 SVM TCGA TC TME TNC TNR TNXB TNW tSNE TPKM XGB Generalized Additive Models Gene Expression Omnibus Generalized Linear Model Gene Ontology Human Epidermal Growth Factor Receptor 2 Heparan Sulfate Proteoglycan Immunohistochemistry Inflammatory CAFs

FAQs

What was the main goal of this study on breast cancer?

To identify and validate refined marker genes for cancer-associated fibroblasts (CAFs) and to develop a diagnostic model that improves breast cancer diagnosis and guides therapeutic strategies.

Which machine learning approach performed best in building the diagnostic model?

Among the evaluated algorithms, Random Forest showed the best performance due to robust classification accuracy and stability.

How did the study validate the proposed CAF biomarkers?

It used immunohistochemical (IHC) experiments to verify expression patterns of the biomarkers, including FXYD1, SULF1, and TNXB, and compared normal versus cancerous tissues.

Sep 22, 2026

Upload by Patrick