A Macro–micro–macro Radiogenomic Framework Identifies FIBCD1 as a Key Immune-modulating Biomarker in Breast Cancer

A Macro–micro–macro Radiogenomic Framework Identifies FIBCD1 as a Key Immune-modulating Biomarker in Breast Cancer | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article A Macro–micro–macro Radiogenomic Framework Identifies FIBCD1 as a Key Immune-modulating Biomarker in Breast Cancer Minping Hong, Keng Ling, Xiaobo Chen, Xiaowen Huang, Wenjing Jiang, and 6 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7024804/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 25 Nov, 2025 Read the published version in Journal of Translational Medicine → Version 1 posted 4 You are reading this latest preprint version Abstract Background Breast cancer prognosis remains challenging due to tumor heterogeneity and the limited predictive power of conventional clinical models. Integrating imaging features with molecular data may improve individualized risk stratification and clinical decision-making. Methods We developed a closed-loop prognostic model based on a macro–micro–macro radiogenomic framework that combines MRI-based radiomics with transcriptomic and proteomic data. A total of 788 radiomics-guided candidate genes were screened. Prognostic gene signatures were identified using multiple machine learning algorithms and validated in TCGA and GEO cohorts. We further analyzed immune infiltration, drug sensitivity, and gene enrichment profiles across risk groups. Causal relationships between gene expression and survival were assessed using Mendelian randomization. FIBCD1 expression was validated in patient plasma using ELISA, and Olink proteomics and radiomic analyses were conducted for biological interpretation. Results We identified a 10-gene prognostic signature. The combined Elastic Net and stepwise Cox regression model achieved the highest concordance index (C-index = 0.645). High-risk patients showed reduced immune activation, increased expression of pro-inflammatory cytokines such as IL-6, and shorter survival. FIBCD1 emerged as a key prognostic gene with a significant causal association with survival. Its expression was significantly elevated in high-risk plasma samples, positively correlated with inflammatory proteins (e.g., OSM, CASP-8), and associated with specific radiomic features such as tumor sphericity. Conclusions Our findings establish a novel radiogenomic strategy that bridges MRI-derived imaging phenotypes with molecular mechanisms. FIBCD1 may serve as an immune-modulating prognostic biomarker linked to imaging characteristics, providing new insights into non-invasive breast cancer risk assessment and therapeutic targeting. Breast Cancer Radiomics Prognostic Model Transcriptomics FIBCD1 Immune Microenvironment Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction Breast cancer is the most common malignant tumor among women all over the world[1]. Despite considerable progress made in surgical techniques, chemotherapy, radiotherapy and endocrine therapy, the high heterogeneity of breast cancer still poses significant challenges to prognosis assessment and individualized treatment planning[2–4]. Traditional prognostic models for breast cancer usually rely on clinical and pathological features, such as tumor size, lymph node involvement and tumor stage[5–7]. However, these models often fail to fully incorporate molecular characteristics, medical imaging data and other potential biomarkers, thereby limiting the accuracy and stability of their predictions. In recent years, integrating multi-omics data for tumor diagnosis and prognosis prediction has become an emerging trend[8, 9]. In particular, radiomics extracts high-throughput macroscopic tumor features (such as shape, texture and spatial heterogeneity) from medical images[10, 11]. Transcriptomic analysis delves into the microscopic domain to clarify the gene expression patterns and molecular pathways that drive tumorigenesis and development[12].Nevertheless, effectively combining the macroscopic insights of radiomics with the microscopic details of transcriptomics to develop powerful and clinically usable predictive models remains a formidable challenge[13, 14]. In our previous work, we constructed a machine-learning model based on contrast-enhanced MRI radiomics features that successfully predicted the axillary lymph node (ALN) status in patients with breast cancer[15]. From this study, we simultaneously identified multiple sets of candidate genes significantly associated with relevant radiomic traits. Given the close relationship between ALN status and patient survival or relapse risk[16–18], these candidate genes potentially offer valuable insights for subsequent prognostic model development. Building on this foundation, we selected 788 candidate genes guided by radiomic analyses and aimed to construct and validate a xcomprehensive, accurate prognostic model for breast cancer[15]. To overcome the limitations of single-layer data modeling, we developed a closed-loop analytical pipeline integrating radiomics-guided gene screening, multi-omics modeling, and biological validation. This strategy aims to bridge macroscopic imaging features with microscopic molecular mechanisms, thereby enhancing the model's explanatory power and potential for clinical translation. Furthermore, the introduction of Mendelian Randomization enhances the causal reasoning between hub genes and clinical outcomes at the genetic level[11, 19, 20], further increasing the robustness of the model. Materials and Methods Data Sources The overall workflow was detailed in Fig. 1 . Gene expression data were obtained from The Cancer Genome Atlas (TCGA) Breast Cancer dataset (BRCA), and the Gene Expression Omnibus (GEO) database, specifically dataset GSE20685. The TCGA dataset comprises 1,118 breast cancer samples and 113 normal controls, while the GEO dataset (GSE20685) includes 327 breast cancer samples. The Mendelian Randomization analysis data from the eQTLGen consortium (Breast_Mammary_Tissue.lite) and FinnGen’s genome-wide association study (GWAS) of breast cancer. The set of 788 candidate genes utilized in this study was previously identified through radiomics and genomic association analyses of breast cancer MRI data, as reported in our earlier publication[15]. These genes were found to be significantly associated with ALN metastasis. The peripheral blood samples used for ELISA and Olink analysis were collected from breast cancer patients in Jiaxing TCM Hospital Affiliated to Zhejiang Chinese Medical University, comprising low ALN burden (n = 40) and high ALN burden (n = 40). This study was approved by the hospital ethics committee and informed consent was obtained from all patients. Additionally, 117 breast MRI scans from the TCGA database were utilized to assess the correlation between hub genes and radiomic features.​ Differential Expression and Selection of Prognostic Genes Differentially expressed genes (DEGs) between breast cancer and normal tissues were found in the TCGA and GEO datasets. Subsequently, through univariate Cox regression analysis, genes significantly associated with overall survival (OS) were screened. Construction of the Machine Learning Models Based on the prognostically relevant genes, multiple machine learning approaches—including Random Survival Forest, Elastic Net, Lasso regression, Ridge regression, stepwise Cox regression, CoxBoost, partial least squares Cox regression, supervised principal component analysis, generalized boosted regression modeling, and survival support vector machines—were employed to build prognostic models. Prior to model construction, gene expression data from different datasets were merged and subjected to batch effect correction using the ComBat algorithm to minimize technical variability. A leave-one-out cross-validation (LOOCV) framework was utilized, with the TCGA-BRCA cohort as the training set and GSE20685 as the external validation set, resulting in a total of 101 model combinations. Each model’s predictive performance was evaluated via concordance index (C-index) on the validation data, and a risk score (RiskScore) was calculated based on the linear combination of gene expression levels. The model with the highest average C-index was deemed optimal. Model Validation After identifying the optimal model, we used the median RiskScore in the TCGA training set as a cutoff to classify patients into high- and low-risk groups. Kaplan - Meier survival curves and log-rank tests were performed (using the “survival” and “survminer” packages in R) to compare survival differences between groups. Subsequently, univariate and multivariate Cox regression analyses were conducted to evaluate the independent prognostic value of the RiskScore, and a nomogram was constructed to visually represent the model’s predictive performance. Immune Infiltration and Tumor Microenvironment Analysis To explore immune microenvironment differences between two groups, seven algorithms were used to quantify immune cell infiltration in the breast cancer samples, and Pearson correlation analysis was conducted to assess the relationship between RiskScore and immune cell abundance. Additionally, based on the immune subtypes defined by Thorsson et al.[21], the “estimate” R package was employed to calculate tumor microenvironment scores, and immune checkpoint expression was compared across high- and low-risk groups. Gene Set Enrichment Analysis (GSEA) Specifically, gene set enrichment analysis was used to identify tumor microenvironment related distinct biological processes for RNA-based risk stratification. The Gene Ontology gene set pathway from the Molecular Signatures Database ( https://www.gsea-msigdb.org/ ) were analyzed within this workflow. Significant pathways were determined according to the absolute value of normalized enrichment score > 1, P < 0.05, and Benjamini-Hochberg adjust p value < 0.25 in TCGA cohort. Drug Sensitivity Analysis Data from the Genomics of Drug Sensitivity in Cancer 2 (GDSC2) repository were utilized with the “oncoPredict” R package to investigate correlations between an RiskScore and the half-maximal inhibitory concentration (IC50) of certain anticancer agents. Mendelian Randomization Analysis To examine the causal relationship between gene expression and breast cancer prognosis, summary-data–based Mendelian Randomization analysis was applied. Summary statistics were obtained from the eQTLGen consortium and FinnGen’s genome-wide association study (GWAS) of breast cancer. After harmonizing allelic information, SMR (version 1.3.1) was performed with a significance threshold of p 0.5, and HEIDI p > 0.05, using genetic variants within ± 1,000 kb of each gene (cis region) to assess associations. Plasma Sample Collection and ELISA Validation of Hub Gene Expression Following the identification of the hub gene through SMR, the Kaplan-Meier analysis was performed using the Kaplan-Meier Plotter website ( https://kmplot.com/analysis/ ). Then ELISA assays to validate its protein expression levels in plasma samples. The study cohort included breast cancer patients with ALN high burden and matched controls with low burden. Peripheral blood samples were collected from all participants, and plasma was separated by centrifugation at 4,000 rpm for 10 minutes at room temperature. Protein concentrations of the hub gene in plasma were quantified using a commercially available ELISA kit (Lun Chang Shuo, NoED-202923), with reactions performed at 37°C. A second-order polynomial standard curve was generated by plotting OD values at 450 nm against known concentrations, enabling accurate quantification of unknown samples. For each test, 10 µL of plasma and 40 µL of diluent were added to the substrate solution prior to OD measurements. According to the manufacturer's instructions, each test is repeated three times. Olink Proximity Extension Assay To elucidate the immunological mechanisms associated with the hub gene, we employed the Olink Proximity Extension Assay (PEA) technology to profile plasma protein expression. Serum proteins were analysed using the Olink Target 96 Inflammation panels (Olink Proteomics, Uppsla, Sweden), which comprised 92 proteins uses the methodology based on the proximity extension assay. Data are reported as normalized protein expression (NPX) unit. Detailed experimental procedures and data analysis methods are provided in the Supplementary I. Radiomics Correlation Analysis Next, dynamic contrast-enhanced (DCE) MRI scans were downloaded from the TCGA imaging database. The U2Net deep learning architecture was utilized for automatic volume-of-interest segmentation of the peak-phase DCE-MRI images, and two experienced radiologists (M.P.H. and J.Y.W., 10 years and 22 years of experience) corrected the segmented volumes. To mitigate discrepancies across different MRI scanners, all images were resampled to a uniform voxel spacing of 1 × 1 × 1 mm 3 (x, y, z) using linear interpolation. The tumor region plus an additional 15-pixel three-dimensional margin were segmented, followed by histogram equalization based on field strength intensity. PyRadiomics (version 3.0.1) was employed to extract radiomic features such as shape, first-order, texture, wavelet, exponential, and square transforms. Identical segmentation and feature extraction procedures were repeated in 50 cases after two months to assess intraclass correlation coefficients (ICC). Finally, Pearson correlation analysis was conducted between hub gene expression and radiomic features of interest to identify potential imaging biomarkers.The parameters of breast MRI collection and scanning of the TCIA are detailed in the Supplementary II. Statistical Analysis All data processing, statistical analyses, and plotting were performed in R (version 4.4.1). The Pearson correlation coefficient was used for continuous variables. The Wilcoxon test was used for comparison between low and high risk groups. Kaplan–Meier test and log-rank tests were used for survival analysis, and Cox regression was used to determine prognostic factors. For ELISA assays, statistical significance was assessed using one-way analysis of variance (ANOVA) followed by Tukey’s post-hoc test. All tests were two-sided tests. 0.05 is considered statistically significant. Results Construction and Validation of the Prognostic Model Based on univariate Cox regression analysis (p < 0.05), we initially identified 13 genes with significant associations to OS from the candidate genes linked to ALN burden (Fig. 2 B). In the TCGA-BRCA cohort, 101 different prognostic model combinations were subsequently fitted and then validated in the GSE20685 dataset. Comparison of the C-index for each model on the validation set revealed that the stepwise Cox regression (backward elimination) combined with Elastic Net (α = 0.3) achieved the highest average C-index of 0.645 (Fig. 2 A). From this optimal model, a 10-gene signature was chosen to compute a risk score (RiskScore) (Fig. 2 C). Kaplan-Meier analysis showed that in the TCGA-BRCA training set, the OS of patients in the low-risk group was significantly higher than that in the high-risk group (p < 0.001, Fig. 2 D), this trend was also verified in the external cohort of GSE20685 (Fig. 2 E). Time-dependent ROC analysis showed that 1-year, 3-year, and 5-year OS AUC values of 0.686, 0.739, and 0.689, respectively (Fig. 2 F). The results of Univariate Cox regression showed that RiskScore, age, and T stage were significantly associated with OS (Fig. 2 H), while Multivariate Cox regression analysis confirmed that RiskScore, age and T stage could independently predict the prognosis of patients (p < 0.001, Fig. 2 I). In addition, time-dependent C-index analysis showed that RiskScore outperformed traditional clinical parameters (Fig. 2 J). Nomogram integrating RiskScore, age, and T stage for individualized OS prediction, with AUC values of 0.859, 0.821, and 0.768, respectively (Fig. 2 G, 2 K). Transcriptomic Subtypes and Immune Infiltration–Related Biological Differences Based on seven immune-infiltration algorithms, RiskScore was found to be positively correlated with Common Lymphoid Progenitors (XCELL) and negatively correlated with T cells (MCPCOUNTER). Figure 2 A displays only the immune cell types significantly associated with RiskScore (P < 0.05). Immune subtyping (Fig. 3 B) showed that low-risk patients more frequently exhibited IFN-γ–dominant immune phenotypes, whereas high-risk patients more commonly had subtypes characterized by elevated proinflammatory factors (e.g., IL-6). Further tumor microenvironment scoring (StromalScore, ImmuneScore, ESTIMATEScore) revealed that the high-risk group was significantly lower than the low-risk group (Fig. 3 C, all p < 0.05). Among immune checkpoints, 39 genes differed significantly between high- and low-risk cohorts (Fig. 3 E). Tumor Immune Dysfunction and Exclusion (TIDE) analysis showed that the higher TIDE scores associated with the high-risk group, indicating a poorer response to immunotherapy in the high-risk group (Fig. 3 D). GO enrichment analysis revealed that the low-risk group was significantly enriched in immune-related terms, most notably GOBP_IMMUNOGLOBULIN_PRODUCTION, reflecting enhanced adaptive immune activity (Fig. 3 F). In contrast, the high-risk group showed enrichment in transcription-related terms such as GOCC_NUCLEOSOME, indicative of increased chromatin remodeling and transcriptional activity (Fig. 3 G). Association of RiskScore with Drug Sensitivity In the drug sensitivity analysis, we identified several chemotherapeutic agents whose predicted IC50 values were significantly correlated with RiskScore stratification (p < 0.001). Notably, Palbociclib, Ribociclib, and Niraparib—agents widely used or studied in the treatment of breast cancer—exhibited markedly higher IC50 values in high-risk patients, indicating potential resistance (Fig. 3 H-J). FIBCD1 as a Hub Gene According to Mendelian Randomization analysis, FIBCD1 was identified as a potential risk gene associated with the prognosis of breast cancer (Fig. 4 A-B). Kaplan–Meier survival curves further confirmed that low FIBCD1 expression was correlated with longer OS (Log-rank p = 0.03, Fig. 4 C). Furthermore, ELISA results demonstrated that the expression level of FIBCD1 protein in the peripheral blood of patients with a ALN-high burden was elevated (p < 0.01, Fig. 4 D), confirming that FIBCD1 is also a potential risk gene for ALN metastasis. Associations between Proteomics and FIBCD1 From the peripheral blood Olink Inflammation panels data, 75 immune-related features were extracted. Pearson correlation analysis showed that FIBCD1 expression was positively correlated with OSM (cor = 0.69, p = 1.87E-12), TNFSF14 (cor = 0.62, p = 6.61E-10), CASP-8 (cor = 0.61, p = 1.81E-09), and TGF-alpha (cor = 0.59, p = 4.22E-09), but negatively correlated with CCL11 (cor = − 0.40, p = 0.0002) and MCP-4 (cor = − 0.34, p = 0.0002) (Fig. 4 E). Associations between Radiomics and FIBCD1 From each volume of interest (VOI) in MRI, 944 radiomic features were extracted. Spearman correlation analysis revealed that FIBCD1 expression was positively associated with original_shape_Sphericity (Cor = 0.31, p < 0.05), a metric quantifying the roundness of the tumor shape, and negatively correlated with wavelet.LLH_glcm_Idn (Cor = -0.27, p < 0.05), which measures the local homogeneity of gray-level intensities within the tumor texture. These radiomic features notably overlap with the key predictors previously utilized in our MRI-based model for ALN burden (Fig. 4 F). Discussion In this study, we developed a macro–micro–macro radiogenomic framework that links MRI-derived imaging phenotypes to immune-modulating molecular mechanisms in breast cancer. Through radiomics-guided gene screening and multi-omics validation, we identified FIBCD1 as a central prognostic biomarker associated with immune suppression and adverse clinical outcomes. Notably, this study ultimately identified FIBCD1 as a hub gene highly associated with poor prognosis in breast cancer and potentially a key biological marker influencing tumor imaging phenotypes. Under the guidance of radiomics, the performance of our prognostic model is comparable to other studies[11, 19, 20]. However, unlike previous research, our study provides biological explanations for the relationship between prognosis and underlying molecular mechanisms[11, 19, 20]. We found that patients in the high-risk group often exhibited typical features of chronic inflammation activation and immune evasion, with significant upregulation of inflammatory cytokines such as IL-6, indicating weakened immune responses and reduced anti-tumor capabilities[22, 23]. In contrast, the low-risk group demonstrated an IFN-γ-dominated immune profile with enhanced T-cell activity, which aligns with better prognoses[24, 25]. Drug sensitivity analysis further revealed that high-risk patients exhibited potential resistance to targeted drugs such as Palbociclib, Ribociclib, and Niraparib, suggesting that alternative therapeutic strategies should be designed for this subgroup[26]. Additionally, GSEA enrichment analysis showed that immune pathways were significantly enriched in the low-risk group, whereas metabolic-related pathways were upregulated in the high-risk group, suggesting that metabolic reprogramming may be another potential mechanism influencing prognosis[27]. From a clinical perspective, combined immune modulation and metabolic-targeted strategies may provide a more effective treatment approach for high-risk patients[28]. At the micro-level, we identified FIBCD1 as a hub driver of high-risk scores. The prognostic value of this gene is supported by multiple lines of evidence: First, Mendelian randomization analysis demonstrated a causal relationship between FIBCD1 and breast cancer prognosis, suggesting that its expression is genetically regulated and directly impacts survival outcomes[29]. Second, ELISA experiments validated that FIBCD1 levels were significantly elevated in the peripheral blood of patients with high axillary lymph node (ALN) burden. Third, Olink proteomics analysis showed that high FIBCD1 expression was positively correlated with various pro-inflammatory and immune-suppressive proteins (e.g., TGF-α, CASP-8, EN-RAGE), while it was negatively correlated with chemokines (e.g., CCL11, MCP-4). These findings suggest that FIBCD1 may contribute to the establishment of an "immune cold tumor" phenotype by inhibiting immune cell recruitment and weakening immune responses, thereby promoting immune evasion and tumor progression[30]. Of greater clinical translational significance, we achieved a closed-loop analysis from molecular mechanisms to imaging features at the macro level. Imaging-molecular correlation analysis revealed that FIBCD1 expression was positively correlated with tumor sphericity (original_shape_Sphericity) and negatively correlated with texture signal homogeneity (wavelet.LLH_glcm_Idn). Specifically, high-expression patients tended to exhibit mass-like tumors with higher sphericity and greater textural heterogeneity, which are the hub radiomic features we identified in the ALN prediction model[15]. We further validated this imaging-molecular-clinical consistency through two representative cases (Fig. 5 ), demonstrating that FIBCD1 is not only a potential molecular biomarker but also likely plays a role in regulating tumor morphology, thus influencing the formation of imaging features and the accuracy of prognostic predictions[10]. The principal innovation of this study is the development of a macro–micro–macro closed-loop framework that integrates imaging feature–guided gene screening with molecular validation, thereby enhancing the biological interpretability and clinical applicability of breast cancer prognostic assessment. Compared to prior work that relies solely on either imaging or gene expression features, our radiomic-guided gene selection of candidate genes and in-depth immune-microenvironment analysis may prove especially advantageous[31–33]. Moreover, the introduction of Mendelian randomization helps mitigate confounding factors, thus adding further weight to the causal association between FIBCD1 and breast cancer prognosis[34]. This study has certain limitations. First, our study primarily utilized publicly available datasets, which may introduce selection bias and limit the generalizability of the results. Second, while we identified associations between FIBCD1 expression and imaging features, the underlying biological mechanisms remain to be elucidated through functional experiments. Third, the study's retrospective nature necessitates prospective validation in independent cohorts to confirm the clinical utility of the prognostic model. In summary, we developed a precision prognostic model for breast cancer through a closed-loop macro–micro–macro framework, integrating radiomics-guided gene selection, transcriptomic data, and biological validation. This model elucidates FIBCD1’s role in tumor progression and highlights the influence of the immune microenvironment on patient outcomes. Notably, the significant correlation between FIBCD1 expression and key radiomic features provides novel evidence linking imaging phenotypes to underlying molecular biology. Our findings not only offer new perspectives for individualized treatment strategies but also establish a foundation for further exploration of tumor metastasis mechanisms and immunotherapeutic targets. Abbreviations ALN Axillary Lymph Node AUC Area Under the Curve C-index Concordance Index DEG Differentially Expressed Gene DCE-MRI Dynamic Contrast-Enhanced Magnetic Resonance Imaging FDR False Discovery Rate GSEA Gene Set Enrichment Analysis GWAS Genome-Wide Association Study HEIDI Heterogeneity in Dependent Instruments IC50 Half-Maximal Inhibitory Concentration ICC Intraclass Correlation Coefficient NPX Normalized Protein Expression OS Overall Survival OSM Oncostatin M PEA Proximity Extension Assay ROC Receiver Operating Characteristic SMR Summary-data-based Mendelian Randomization TIDE Tumor Immune Dysfunction and Exclusion TME Tumor Microenvironment VOI Volume of Interest Declarations Ethical Approval and Consent to participate This study was approved by the Institutional Review Board of Jiaxing TCM Hospital Affiliated to Zhejiang Chinese Medical University (IRB No: SL-2022-0369, Approval Date: October 18, 2022). All participants provided written informed consent prior to inclusion in the study. Consent for publication All authors have read and approved the final manuscript and consent to its publication. Competing interests The authors declare that they have no competing interests. Acknowledgements The authors gratefully acknowledge the support of the Guangdong Provincial Key Laboratory of Artificial Intelligence in Medical Image Analysis and Application for providing technical guidance and infrastructure during the course of this study. Authors' contributions M.H. , K.L., and H.S. conceived and designed the study; K.L., Z.M., X.H. and X.C. acquired the data; W.J. and J.H. did the statistical analyses; Y.Y. and X.H. implemented quality control of data and the algorithms; All authors analyzed and interpreted the data; K.L., Z.M., X.C. and H.S.. prepared the first draft of the manuscript; M.H. revised the manuscript; All authors contributed to manuscript preparation. Funding The research was supported by the grants of Zhejiang Provincial Natural Science Foundation of China (Grant No. LQN25H180008), Medical Science and Technology Project of Zhejiang Province (Grant No. 2025KY366, 2024KY454, 2025KY1626), Public Welfare Research Project of Jiaxing (Grant No. 2024AY10026), Zhejiang Traditional Chinese Medicine Administration (Grant No. 2024ZL1058) and Shanghai Municipal Health Commission Health Industry Clinical Research Special Program (Youth Project) (20244Y0003). Availability of data and materials The datasets generated and/or analyzed during the current study have been registered with the Chinese Clinical Trial Registry (ChiCTR2500103228). While the data are not publicly available at the time of submission, they will be uploaded to the ChiCTR platform after project completion. During the interim period, data are available from the corresponding author upon reasonable request. Additionally, the code used in the analyses is available on GitHub at https://github.com/idcast-Tracy/Image-genomics. References Siegel RL, Kratzer TB, Giaquinto AN, Sung H, Jemal A. Cancer statistics, 2025. CA Cancer J Clin. 2025;75:10–45. Wang G, Wang S, Song W, Lu C, Chen Z, He L, et al. Integrating multi-omics data reveals the antitumor role and clinical benefits of gamma-delta T cells in triple-negative breast cancer. BMC Cancer. 2025;25:623. Meattini I, Becherini C, Caini S, Coles CE, Cortes J, Curigliano G, et al. International multidisciplinary consensus on the integration of radiotherapy with new systemic treatments for breast cancer: European Society for Radiotherapy and Oncology (ESTRO)-endorsed recommendations. Lancet Oncol. 2024;25:e73–83. Rayson VC, Harris MA, Savas P, Hun ML, Virassamy B, Salgado R, et al. The anti-cancer immune response in breast cancer: current and emerging biomarkers and treatments. Trends Cancer. 2024;10:490–506. Teng J, Zhang H, Liu W, Shu X-O, Ye F. A Dynamic Bayesian Model for Breast Cancer Survival Prediction. IEEE J Biomed Health Inform. 2022;26:5716–27. Wang R, Zhu Y, Liu X, Liao X, He J, Niu L. The Clinicopathological features and survival outcomes of patients with different metastatic sites in stage IV breast cancer. BMC Cancer. 2019;19:1091. Fan R, Chen Y, Nechuta S, Cai H, Gu K, Shi L, et al. Prediction models for breast cancer prognosis among Asian women. Cancer. 2021;127:1758–69. Jiang Y-Z, Ma D, Jin X, Xiao Y, Yu Y, Shi J, et al. Integrated multiomic profiling of breast cancer in the Chinese population reveals patient stratification and therapeutic vulnerabilities. Nat Cancer. 2024;5:673–90. Sammut S-J, Crispin-Ortuzar M, Chin S-F, Provenzano E, Bardwell HA, Ma W, et al. Multi-omic machine learning predictor of breast cancer therapy response. Nature. 2022;601:623–9. Qi Y-J, Su G-H, You C, Zhang X, Xiao Y, Jiang Y-Z, et al. Radiomics in breast cancer: Current advances and future directions. Cell Rep Med. 2024;5:101719. Wu X, Lu X, Zhang W, Zhong X, Bu H, Zhang Z. Development and validation of a 10-gene signature for predicting recurrence risk in HR+/HER2- early breast cancer undergoing chemo-endocrine therapy. Breast. 2025;82:104484. Verma VK, Beevi SS, Nair RA, Kumar A, Kiran R, Alexander LE, et al. MicroRNA signatures differentiate types, grades, and stages of breast invasive ductal carcinoma (IDC): miRNA-target interacting signaling pathways. Cell Commun Signal. 2024;22:100. Chen Z, Yang C, Wang J, Wang B, He L, Liu Z, et al. Integrating of Radiomics and Single-cell Transcriptomics Uncovers the Crucial Role of γδ T Cells in Lung Adenocarcinoma. J Cancer. 2025;16:1167–80. Corti A, Lenoci D, Corino VDA, Mattavelli D, Ravanelli M, Poli T, et al. Interplay between MRI radiomics and immune gene expression signatures in oral squamous cell carcinoma. Sci Rep. 2025;15:12622. Hong M, Fan S, Xu Z, Fang Z, Ling K, Lai P, et al. MRI radiomics and biological correlations for predicting axillary lymph node burden in early-stage breast cancer. J Transl Med. 2024;22:826. Li Z, Gao J, Zhou H, Li X, Zheng T, Lin F, et al. Multiregional dynamic contrast-enhanced MRI-based integrated system for predicting pathological complete response of axillary lymph node to neoadjuvant chemotherapy in breast cancer: multicentre study. EBioMedicine. 2024;107:105311. Kimbung S, Loman N, Hedenfalk I. Clinical and molecular complexity of breast cancer metastases. Semin Cancer Biol. 2015;35:85–95. Schnitt SJ. Classification and prognosis of invasive breast cancer: from morphology to molecular taxonomy. Mod Pathol. 2010;23 Suppl 2:S60-64. Guan J, Fan M, Li L. MVNMF: Multiview nonnegative matrix factorization for radio-multigenomic analysis in breast cancer prognosis. Med Image Anal. 2025;103:103566. Wu X, Li Y, Chen J, Chen J, Zhang W, Lu X, et al. Multimodal recurrence risk prediction model for HR+/HER2- early breast cancer following adjuvant chemo-endocrine therapy: integrating pathology image and clinicalpathological features. Breast Cancer Res. 2025;27:27. Thorsson V, Gibbs DL, Brown SD, Wolf D, Bortone DS, Ou Yang T-H, et al. The Immune Landscape of Cancer. Immunity. 2018;48:812-830.e14. Manore SG, Doheny DL, Wong GL, Lo H-W. IL-6/JAK/STAT3 Signaling in Breast Cancer Metastasis: Biology and Treatment. Front Oncol. 2022;12:866014. Weng Y-S, Tseng H-Y, Chen Y-A, Shen P-C, Al Haq AT, Chen L-M, et al. MCT-1/miR-34a/IL-6/IL-6R signaling axis promotes EMT progression, cancer stemness and M2 macrophage polarization in triple-negative breast cancer. Mol Cancer. 2019;18:42. Cai J, Zhong M, Xu J, Cheng H, Xu S. Codelivery of triptolide and IFN-γ to boost antitumor immunity for triple-negative breast cancer. Int Immunopharmacol. 2023;120:110346. Castro F, Pinto ML, Leite Pereira C, Serre K, Costa ÂM, Cavadas B, et al. Chitosan/γ-PGA nanoparticles and IFN-γ immunotherapy: A dual approach for triple-negative breast cancer treatment. J Control Release. 2025;379:621–35. Li S, Zhang Y, Wang N, Guo R, Liu Q, Lv C, et al. Pan-cancer analysis reveals synergistic effects of CDK4/6i and PARPi combination treatment in RB-proficient and RB-deficient breast cancer cells. Cell Death Dis. 2020;11:219. Ren W, Yu Y, Wang T, Wang X, Su K, Wang Y, et al. Comprehensive analysis of metabolism-related gene biomarkers reveals their impact on the diagnosis and prognosis of triple-negative breast cancer. BMC Cancer. 2025;25:668. Xia L, Oyang L, Lin J, Tan S, Han Y, Wu N, et al. The cancer metabolic reprogramming and immune response. Mol Cancer. 2021;20:28. Eem M, S le C. Instrumental Variables Analysis and Mendelian Randomization for Causal Inference. The Journal of infectious diseases. 2025;231. Galon J, Bruni D. Approaches to treat immune hot, altered and cold tumours with combination immunotherapies. Nat Rev Drug Discov. 2019;18:197–218. Tao Y, Wang Q, Guo S, Liu J, Cao Y. m6A related metabolic genes in breast cancer and their relationship with prognosis. Int Immunopharmacol. 2025;148:114121. Huang R, Li Y, Lin K, Zheng L, Zhu X, Huang L, et al. A novel glycolysis-related gene signature for predicting prognosis and immunotherapy efficacy in breast cancer. Front Immunol. 2025;16:1512859. Wenwen null, Jiang Z, Liu J, Liu D, Li Y, He Y, et al. Integrating ultrasound radiomics and clinicopathological features for machine learning-based survival prediction in patients with nonmetastatic triple-negative breast cancer. BMC Cancer. 2025;25:291. Evans DM, Davey Smith G. Mendelian Randomization: New Applications in the Coming Age of Hypothesis-Free Causality. Annu Rev Genomics Hum Genet. 2015;16:327–50. Supplementary Files Supplementarymaterial.docx Cite Share Download PDF Status: Published Journal Publication published 25 Nov, 2025 Read the published version in Journal of Translational Medicine → Version 1 posted Reviewers agreed at journal 03 Aug, 2025 Reviewers invited by journal 03 Aug, 2025 Editor assigned by journal 04 Jul, 2025 First submitted to journal 01 Jul, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7024804","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":494949478,"identity":"88fd767d-62ef-4430-bb7d-2ee2484be821","order_by":0,"name":"Minping Hong","email":"","orcid":"","institution":"Zhejiang Chinese Medical University","correspondingAuthor":false,"prefix":"","firstName":"Minping","middleName":"","lastName":"Hong","suffix":""},{"id":494949479,"identity":"91cbfcf0-82ea-4688-b95c-c26ca7b2efd6","order_by":1,"name":"Keng Ling","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAzUlEQVRIiWNgGAWjYBACxvkHGx98MLCx42dvIFIL8wzmw4YzKtKSJXsOEKmFfQZbmjDPmcOMG24kEKmFd3aPGQNvGzOzwc3HG28w1NhEE9QiOeeM2QPJNjY+ydtpxRYMx9JyGwhpMWzIMTcwbONh5rudYybB2HCYsBb7A0CViW1AxTfPEKmFcUZamsSBMwaME27wEKul5/Bhw4aKBGAgA/2SQIxfGNsbGx//MfgPjMrDG298qLEhrAUZGEgkkKIcooVUHaNgFIyCUTAyAADct0PrUjN8cQAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0003-0081-5069","institution":"Jiaxing Maternity and Children Health Care Hospital","correspondingAuthor":true,"prefix":"","firstName":"Keng","middleName":"","lastName":"Ling","suffix":""},{"id":494949480,"identity":"8ba2f6b6-e422-48cd-bff8-35bd2dcbe0ec","order_by":2,"name":"Xiaobo Chen","email":"","orcid":"","institution":"Guangdong Provincial People's Hospital Affiliated to Southern Medical University: Guangdong Provincial People's Hospital","correspondingAuthor":false,"prefix":"","firstName":"Xiaobo","middleName":"","lastName":"Chen","suffix":""},{"id":494949481,"identity":"413c30ab-1ccb-4a69-9f06-5902e537ebc0","order_by":3,"name":"Xiaowen Huang","email":"","orcid":"","institution":"Zhejiang Chinese Medical University","correspondingAuthor":false,"prefix":"","firstName":"Xiaowen","middleName":"","lastName":"Huang","suffix":""},{"id":494949482,"identity":"e3820ff5-efa6-4890-b316-4f6b13f80c03","order_by":4,"name":"Wenjing Jiang","email":"","orcid":"","institution":"Guangdong Provincial People\\'s Hospital: Guangdong Provincial People's Hospital","correspondingAuthor":false,"prefix":"","firstName":"Wenjing","middleName":"","lastName":"Jiang","suffix":""},{"id":494949483,"identity":"4f8fbbe5-dc1a-44ac-afab-0bf3544c8b80","order_by":5,"name":"Jie He","email":"","orcid":"","institution":"Zhejiang University","correspondingAuthor":false,"prefix":"","firstName":"Jie","middleName":"","lastName":"He","suffix":""},{"id":494949484,"identity":"6d6f2a15-c963-480e-88c5-c8ca81ceb1e7","order_by":6,"name":"Jie Hou","email":"","orcid":"","institution":"Zhejiang Provincial People's Hospital","correspondingAuthor":false,"prefix":"","firstName":"Jie","middleName":"","lastName":"Hou","suffix":""},{"id":494949485,"identity":"497f9415-0844-42ef-8035-ee940ef36fdf","order_by":7,"name":"Yujie Ying","email":"","orcid":"","institution":"Zhejiang Chinese Medical University","correspondingAuthor":false,"prefix":"","firstName":"Yujie","middleName":"","lastName":"Ying","suffix":""},{"id":494949486,"identity":"ed82d8f4-0e64-4f85-85e8-8fdd0e3d4c8f","order_by":8,"name":"Bing Zhou","email":"","orcid":"","institution":"Zhejiang Chinese Medical University","correspondingAuthor":false,"prefix":"","firstName":"Bing","middleName":"","lastName":"Zhou","suffix":""},{"id":494949487,"identity":"b3653490-f77c-4e1a-9817-15a7b31356f2","order_by":9,"name":"Zhenyi Ma","email":"","orcid":"","institution":"Zhejiang Chinese Medical University","correspondingAuthor":false,"prefix":"","firstName":"Zhenyi","middleName":"","lastName":"Ma","suffix":""},{"id":494949488,"identity":"c8227d5e-096e-4a99-be73-fd70d8b5bf4c","order_by":10,"name":"Haitao Sun","email":"","orcid":"","institution":"Zhongshan Hospital Fudan University","correspondingAuthor":false,"prefix":"","firstName":"Haitao","middleName":"","lastName":"Sun","suffix":""}],"badges":[],"createdAt":"2025-07-02 03:45:54","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7024804/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7024804/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1186/s12967-025-07389-z","type":"published","date":"2025-11-25T15:58:19+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":88523236,"identity":"6a1ff5e4-12f5-45a9-b1f2-cc8cb5382519","added_by":"auto","created_at":"2025-08-07 10:02:28","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":971859,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFlowchart.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMendelian Randomization (MR), a statistical method used to validate causal relationships between the identified hub gene and breast cancer prognosis. This step helps establish the genetic basis underlying the association between hub gene expression and poor clinical outcomes.\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-7024804/v1/c52e2b1320a3898647e99676.png"},{"id":88523237,"identity":"404ddbf6-d422-47f7-abe5-3257eb8fca13","added_by":"auto","created_at":"2025-08-07 10:02:28","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":7508993,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePrognostic model construction and validation.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(A) Bar plot shows comparison of C-index values for 101 model combinations in the TCGA-BRCA and GSE20685 cohorts. (B) Univariate Cox regression analysis identified 13 genes significantly associated with overall survival (OS). (C) Final 10-gene signature selected from the optimal model to calculate RiskScore. (D-E) Kaplan-Meier survival curves in the TCGA-BRCA training set and external validation cohort GSE20685. (F-G) Time-dependent ROC curves of the RiskScore and the nomogram. (H-I) Forest plots of univariate (H) and multivariate (I) Cox regression analyses for independent prognostic factors. (J) Time-dependent C-index analysis (RiskScore vs. clinical parameters). (K) Nomogram combining RiskScore, age, and T stage for individualized prognosis.\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-7024804/v1/31a3ef19ae8483551b32a658.png"},{"id":88523234,"identity":"9c2f4a8c-2ae0-4ade-a06e-21a31b9e4560","added_by":"auto","created_at":"2025-08-07 10:02:28","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":4489464,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eTranscriptomic subtypes, immune infiltration, and drug sensitivity analysis.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(A) Correlation between RiskScore and immune cell infiltration. (B) Risk Stratification and Subtype Distribution of 937 patients from TCGA (C1: wound healing; C2: IFN-γ dominant; C3: inflammatory; C4: lymphocyte depleted; C6: TGF-β dominant). (C) Tumor microenvironment scores significantly lower in high-risk group. (D) Boxplots show higher TIDE scores in high-risk patients indicate potential immunotherapy resistance. (E) Differential expression of 39 immune checkpoint genes between risk groups. (F-G) Enrichment plots shows Immune-related pathways upregulated in low-risk group (F), metabolic pathways enriched in high-risk group (G). (H-J) Higher IC50 values for Palbociclib (H), Ribociclib (I), and Niraparib (J) in high-risk patients.\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-7024804/v1/9cbf3508e4be0777ea8f4945.png"},{"id":88523241,"identity":"fcb92780-1a34-4e13-acb9-fc88358b341d","added_by":"auto","created_at":"2025-08-07 10:02:28","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":3353070,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eValidation of FIBCD1 as a hub gene and its multi-omics associations.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(A-B) SMR analysis identifies FIBCD1 as a risk gene for breast cancer (ENSG00000130720 represents the FIBCD1 gene). (C) Kaplan-Meier curve shows low FIBCD1 expression correlates with longer overall survival. (D) ELISA analyse shows elevated FIBCD1 expression protein levels in peripheral blood of ALN-high burden patients. (E) FIBCD1-RiskScore correlations. (F) FIBCD1-Radiomic features correlations.\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-7024804/v1/edc45ed3e9f80ba475b186bc.png"},{"id":88523239,"identity":"152896fd-df12-44e7-ac61-c7865e1cbb7d","added_by":"auto","created_at":"2025-08-07 10:02:28","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":1828510,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eRepresentative Cases Illustrating the Macro-Micro-Macro Relationship Among Imaging Features, FIBCD1 Expression, and ALN Burden\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePatient A: \u003c/strong\u003eMRI reveals a mass-like lesion with high sphericity and low texture homogeneity (wavelet.LLH_glcm_Idn). Peripheral blood analysis shows elevated FIBCD1 expression. Clinically, the patient presents with a high axillary lymph node (ALN) burden. \u003cstrong\u003ePatient B: \u003c/strong\u003eMRI reveals a non-mass-like lesion with low sphericity and high texture homogeneity. Peripheral blood analysis shows low FIBCD1 expression. Clinically, the patient presents with a low ALN burden.\u003c/p\u003e","description":"","filename":"Figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-7024804/v1/b82403f909748d86d3a46a14.png"},{"id":97178589,"identity":"79a57cfe-9366-427a-8d8b-3de7fc4578fb","added_by":"auto","created_at":"2025-12-01 16:11:10","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":20029061,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7024804/v1/d55cf748-61f4-4015-8cca-2dce463ca7e1.pdf"},{"id":88524347,"identity":"c368fb49-a90f-44f7-b8f4-f99fd0cdc56e","added_by":"auto","created_at":"2025-08-07 10:10:28","extension":"docx","order_by":9,"title":"","display":"","copyAsset":false,"role":"supplement","size":15231,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarymaterial.docx","url":"https://assets-eu.researchsquare.com/files/rs-7024804/v1/39bd45e56b0574a6094f8fc7.docx"}],"financialInterests":"","formattedTitle":"A Macro–micro–macro Radiogenomic Framework Identifies FIBCD1 as a Key Immune-modulating Biomarker in Breast Cancer","fulltext":[{"header":"Introduction","content":"\u003cp\u003eBreast cancer is the most common malignant tumor among women all over the world[1]. Despite considerable progress made in surgical techniques, chemotherapy, radiotherapy and endocrine therapy, the high heterogeneity of breast cancer still poses significant challenges to prognosis assessment and individualized treatment planning[2\u0026ndash;4]. Traditional prognostic models for breast cancer usually rely on clinical and pathological features, such as tumor size, lymph node involvement and tumor stage[5\u0026ndash;7]. However, these models often fail to fully incorporate molecular characteristics, medical imaging data and other potential biomarkers, thereby limiting the accuracy and stability of their predictions.\u003c/p\u003e\u003cp\u003eIn recent years, integrating multi-omics data for tumor diagnosis and prognosis prediction has become an emerging trend[8, 9]. In particular, radiomics extracts high-throughput macroscopic tumor features (such as shape, texture and spatial heterogeneity) from medical images[10, 11]. Transcriptomic analysis delves into the microscopic domain to clarify the gene expression patterns and molecular pathways that drive tumorigenesis and development[12].Nevertheless, effectively combining the macroscopic insights of radiomics with the microscopic details of transcriptomics to develop powerful and clinically usable predictive models remains a formidable challenge[13, 14].\u003c/p\u003e\u003cp\u003eIn our previous work, we constructed a machine-learning model based on contrast-enhanced MRI radiomics features that successfully predicted the axillary lymph node (ALN) status in patients with breast cancer[15]. From this study, we simultaneously identified multiple sets of candidate genes significantly associated with relevant radiomic traits. Given the close relationship between ALN status and patient survival or relapse risk[16\u0026ndash;18], these candidate genes potentially offer valuable insights for subsequent prognostic model development.\u003c/p\u003e\u003cp\u003eBuilding on this foundation, we selected 788 candidate genes guided by radiomic analyses and aimed to construct and validate a xcomprehensive, accurate prognostic model for breast cancer[15]. To overcome the limitations of single-layer data modeling, we developed a closed-loop analytical pipeline integrating radiomics-guided gene screening, multi-omics modeling, and biological validation. This strategy aims to bridge macroscopic imaging features with microscopic molecular mechanisms, thereby enhancing the model's explanatory power and potential for clinical translation. Furthermore, the introduction of Mendelian Randomization enhances the causal reasoning between hub genes and clinical outcomes at the genetic level[11, 19, 20], further increasing the robustness of the model.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cp\u003e\u003cb\u003eData Sources\u003c/b\u003e\u003c/p\u003e\u003cp\u003eThe overall workflow was detailed in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. Gene expression data were obtained from The Cancer Genome Atlas (TCGA) Breast Cancer dataset (BRCA), and the Gene Expression Omnibus (GEO) database, specifically dataset GSE20685. The TCGA dataset comprises 1,118 breast cancer samples and 113 normal controls, while the GEO dataset (GSE20685) includes 327 breast cancer samples. The Mendelian Randomization analysis data from the eQTLGen consortium (Breast_Mammary_Tissue.lite) and FinnGen\u0026rsquo;s genome-wide association study (GWAS) of breast cancer. The set of 788 candidate genes utilized in this study was previously identified through radiomics and genomic association analyses of breast cancer MRI data, as reported in our earlier publication[15]. These genes were found to be significantly associated with ALN metastasis. The peripheral blood samples used for ELISA and Olink analysis were collected from breast cancer patients in Jiaxing TCM Hospital Affiliated to Zhejiang Chinese Medical University, comprising low ALN burden (n\u0026thinsp;=\u0026thinsp;40) and high ALN burden (n\u0026thinsp;=\u0026thinsp;40). This study was approved by the hospital ethics committee and informed consent was obtained from all patients. Additionally, 117 breast MRI scans from the TCGA database were utilized to assess the correlation between hub genes and radiomic features.​\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eDifferential Expression and Selection of Prognostic Genes\u003c/b\u003e\u003c/p\u003e\u003cp\u003eDifferentially expressed genes (DEGs) between breast cancer and normal tissues were found in the TCGA and GEO datasets. Subsequently, through univariate Cox regression analysis, genes significantly associated with overall survival (OS) were screened.\u003c/p\u003e\u003cp\u003e\u003cb\u003eConstruction of the Machine Learning Models\u003c/b\u003e\u003c/p\u003e\u003cp\u003eBased on the prognostically relevant genes, multiple machine learning approaches\u0026mdash;including Random Survival Forest, Elastic Net, Lasso regression, Ridge regression, stepwise Cox regression, CoxBoost, partial least squares Cox regression, supervised principal component analysis, generalized boosted regression modeling, and survival support vector machines\u0026mdash;were employed to build prognostic models. Prior to model construction, gene expression data from different datasets were merged and subjected to batch effect correction using the ComBat algorithm to minimize technical variability. A leave-one-out cross-validation (LOOCV) framework was utilized, with the TCGA-BRCA cohort as the training set and GSE20685 as the external validation set, resulting in a total of 101 model combinations. Each model\u0026rsquo;s predictive performance was evaluated via concordance index (C-index) on the validation data, and a risk score (RiskScore) was calculated based on the linear combination of gene expression levels. The model with the highest average C-index was deemed optimal.\u003c/p\u003e\u003cp\u003e\u003cb\u003eModel Validation\u003c/b\u003e\u003c/p\u003e\u003cp\u003eAfter identifying the optimal model, we used the median RiskScore in the TCGA training set as a cutoff to classify patients into high- and low-risk groups. Kaplan - Meier survival curves and log-rank tests were performed (using the \u0026ldquo;survival\u0026rdquo; and \u0026ldquo;survminer\u0026rdquo; packages in R) to compare survival differences between groups. Subsequently, univariate and multivariate Cox regression analyses were conducted to evaluate the independent prognostic value of the RiskScore, and a nomogram was constructed to visually represent the model\u0026rsquo;s predictive performance.\u003c/p\u003e\u003cp\u003e\u003cb\u003eImmune Infiltration and Tumor Microenvironment Analysis\u003c/b\u003e\u003c/p\u003e\u003cp\u003eTo explore immune microenvironment differences between two groups, seven algorithms were used to quantify immune cell infiltration in the breast cancer samples, and Pearson correlation analysis was conducted to assess the relationship between RiskScore and immune cell abundance. Additionally, based on the immune subtypes defined by Thorsson et al.[21], the \u0026ldquo;estimate\u0026rdquo; R package was employed to calculate tumor microenvironment scores, and immune checkpoint expression was compared across high- and low-risk groups.\u003c/p\u003e\u003cp\u003e\u003cb\u003eGene Set Enrichment Analysis (GSEA)\u003c/b\u003e\u003c/p\u003e\u003cp\u003eSpecifically, gene set enrichment analysis was used to identify tumor microenvironment related distinct biological processes for RNA-based risk stratification. The Gene Ontology gene set pathway from the Molecular Signatures Database (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.gsea-msigdb.org/\u003c/span\u003e\u003cspan address=\"https://www.gsea-msigdb.org/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) were analyzed within this workflow. Significant pathways were determined according to the absolute value of normalized enrichment score\u0026thinsp;\u0026gt;\u0026thinsp;1, P\u0026thinsp;\u0026lt;\u0026thinsp;0.05, and Benjamini-Hochberg adjust p value\u0026thinsp;\u0026lt;\u0026thinsp;0.25 in TCGA cohort.\u003c/p\u003e\u003cp\u003e\u003cb\u003eDrug Sensitivity Analysis\u003c/b\u003e\u003c/p\u003e\u003cp\u003eData from the Genomics of Drug Sensitivity in Cancer 2 (GDSC2) repository were utilized with the \u0026ldquo;oncoPredict\u0026rdquo; R package to investigate correlations between an RiskScore and the half-maximal inhibitory concentration (IC50) of certain anticancer agents.\u003c/p\u003e\u003cp\u003e\u003cb\u003eMendelian Randomization Analysis\u003c/b\u003e\u003c/p\u003e\u003cp\u003eTo examine the causal relationship between gene expression and breast cancer prognosis, summary-data\u0026ndash;based Mendelian Randomization analysis was applied. Summary statistics were obtained from the eQTLGen consortium and FinnGen\u0026rsquo;s genome-wide association study (GWAS) of breast cancer. After harmonizing allelic information, SMR (version 1.3.1) was performed with a significance threshold of p\u0026thinsp;\u0026lt;\u0026thinsp;0.05, FDR\u0026thinsp;\u0026gt;\u0026thinsp;0.5, and HEIDI p\u0026thinsp;\u0026gt;\u0026thinsp;0.05, using genetic variants within \u0026plusmn;\u0026thinsp;1,000 kb of each gene (cis region) to assess associations.\u003c/p\u003e\u003cp\u003e\u003cb\u003ePlasma Sample Collection and ELISA Validation of Hub Gene Expression\u003c/b\u003e\u003c/p\u003e\u003cp\u003eFollowing the identification of the hub gene through SMR, the Kaplan-Meier analysis was performed using the Kaplan-Meier Plotter website (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://kmplot.com/analysis/\u003c/span\u003e\u003cspan address=\"https://kmplot.com/analysis/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). Then ELISA assays to validate its protein expression levels in plasma samples. The study cohort included breast cancer patients with ALN high burden and matched controls with low burden. Peripheral blood samples were collected from all participants, and plasma was separated by centrifugation at 4,000 rpm for 10 minutes at room temperature. Protein concentrations of the hub gene in plasma were quantified using a commercially available ELISA kit (Lun Chang Shuo, NoED-202923), with reactions performed at 37\u0026deg;C. A second-order polynomial standard curve was generated by plotting OD values at 450 nm against known concentrations, enabling accurate quantification of unknown samples. For each test, 10 \u0026micro;L of plasma and 40 \u0026micro;L of diluent were added to the substrate solution prior to OD measurements. According to the manufacturer's instructions, each test is repeated three times.\u003c/p\u003e\u003cp\u003e\u003cb\u003eOlink Proximity Extension Assay\u003c/b\u003e\u003c/p\u003e\u003cp\u003eTo elucidate the immunological mechanisms associated with the hub gene, we employed the Olink Proximity Extension Assay (PEA) technology to profile plasma protein expression. Serum proteins were analysed using the Olink Target 96 Inflammation panels (Olink Proteomics, Uppsla, Sweden), which comprised 92 proteins uses the methodology based on the proximity extension assay. Data are reported as normalized protein expression (NPX) unit. Detailed experimental procedures and data analysis methods are provided in the Supplementary I.\u003c/p\u003e\u003cp\u003e\u003cb\u003eRadiomics Correlation Analysis\u003c/b\u003e\u003c/p\u003e\u003cp\u003eNext, dynamic contrast-enhanced (DCE) MRI scans were downloaded from the TCGA imaging database. The U2Net deep learning architecture was utilized for automatic volume-of-interest segmentation of the peak-phase DCE-MRI images, and two experienced radiologists (M.P.H. and J.Y.W., 10 years and 22 years of experience) corrected the segmented volumes. To mitigate discrepancies across different MRI scanners, all images were resampled to a uniform voxel spacing of 1 \u0026times; 1 \u0026times; 1 mm\u003csup\u003e3\u003c/sup\u003e (x, y, z) using linear interpolation. The tumor region plus an additional 15-pixel three-dimensional margin were segmented, followed by histogram equalization based on field strength intensity. PyRadiomics (version 3.0.1) was employed to extract radiomic features such as shape, first-order, texture, wavelet, exponential, and square transforms. Identical segmentation and feature extraction procedures were repeated in 50 cases after two months to assess intraclass correlation coefficients (ICC). Finally, Pearson correlation analysis was conducted between hub gene expression and radiomic features of interest to identify potential imaging biomarkers.The parameters of breast MRI collection and scanning of the TCIA are detailed in the Supplementary II.\u003c/p\u003e\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003eStatistical Analysis\u003c/h2\u003e\u003cp\u003eAll data processing, statistical analyses, and plotting were performed in R (version 4.4.1). The Pearson correlation coefficient was used for continuous variables. The Wilcoxon test was used for comparison between low and high risk groups. Kaplan\u0026ndash;Meier test and log-rank tests were used for survival analysis, and Cox regression was used to determine prognostic factors. For ELISA assays, statistical significance was assessed using one-way analysis of variance (ANOVA) followed by Tukey\u0026rsquo;s post-hoc test. All tests were two-sided tests. 0.05 is considered statistically significant.\u003c/p\u003e\u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cb\u003eConstruction and Validation of the Prognostic Model\u003c/b\u003e\u003c/p\u003e\u003cp\u003eBased on univariate Cox regression analysis (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05), we initially identified 13 genes with significant associations to OS from the candidate genes linked to ALN burden (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB). In the TCGA-BRCA cohort, 101 different prognostic model combinations were subsequently fitted and then validated in the GSE20685 dataset. Comparison of the C-index for each model on the validation set revealed that the stepwise Cox regression (backward elimination) combined with Elastic Net (α\u0026thinsp;=\u0026thinsp;0.3) achieved the highest average C-index of 0.645 (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA). From this optimal model, a 10-gene signature was chosen to compute a risk score (RiskScore) (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eC).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eKaplan-Meier analysis showed that in the TCGA-BRCA training set, the OS of patients in the low-risk group was significantly higher than that in the high-risk group (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001, Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eD), this trend was also verified in the external cohort of GSE20685 (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eE). Time-dependent ROC analysis showed that 1-year, 3-year, and 5-year OS AUC values of 0.686, 0.739, and 0.689, respectively (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eF). The results of Univariate Cox regression showed that RiskScore, age, and T stage were significantly associated with OS (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eH), while Multivariate Cox regression analysis confirmed that RiskScore, age and T stage could independently predict the prognosis of patients (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001, Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eI). In addition, time-dependent C-index analysis showed that RiskScore outperformed traditional clinical parameters (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eJ). Nomogram integrating RiskScore, age, and T stage for individualized OS prediction, with AUC values of 0.859, 0.821, and 0.768, respectively (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eG, \u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eK).\u003c/p\u003e\u003cp\u003e\u003cb\u003eTranscriptomic Subtypes and Immune Infiltration\u0026ndash;Related Biological Differences\u003c/b\u003e\u003c/p\u003e\u003cp\u003eBased on seven immune-infiltration algorithms, RiskScore was found to be positively correlated with Common Lymphoid Progenitors (XCELL) and negatively correlated with T cells (MCPCOUNTER). Figure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA displays only the immune cell types significantly associated with RiskScore (P\u0026thinsp;\u0026lt;\u0026thinsp;0.05). Immune subtyping (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB) showed that low-risk patients more frequently exhibited IFN-γ\u0026ndash;dominant immune phenotypes, whereas high-risk patients more commonly had subtypes characterized by elevated proinflammatory factors (e.g., IL-6). Further tumor microenvironment scoring (StromalScore, ImmuneScore, ESTIMATEScore) revealed that the high-risk group was significantly lower than the low-risk group (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC, all p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). Among immune checkpoints, 39 genes differed significantly between high- and low-risk cohorts (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eE). Tumor Immune Dysfunction and Exclusion (TIDE) analysis showed that the higher TIDE scores associated with the high-risk group, indicating a poorer response to immunotherapy in the high-risk group (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eD).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eGO enrichment analysis revealed that the low-risk group was significantly enriched in immune-related terms, most notably GOBP_IMMUNOGLOBULIN_PRODUCTION, reflecting enhanced adaptive immune activity (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eF). In contrast, the high-risk group showed enrichment in transcription-related terms such as GOCC_NUCLEOSOME, indicative of increased chromatin remodeling and transcriptional activity (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eG).\u003c/p\u003e\u003cp\u003e\u003cb\u003eAssociation of RiskScore with Drug Sensitivity\u003c/b\u003e\u003c/p\u003e\u003cp\u003eIn the drug sensitivity analysis, we identified several chemotherapeutic agents whose predicted IC50 values were significantly correlated with RiskScore stratification (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Notably, Palbociclib, Ribociclib, and Niraparib\u0026mdash;agents widely used or studied in the treatment of breast cancer\u0026mdash;exhibited markedly higher IC50 values in high-risk patients, indicating potential resistance (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eH-J).\u003c/p\u003e\u003cp\u003e\u003cb\u003eFIBCD1 as a Hub Gene\u003c/b\u003e\u003c/p\u003e\u003cp\u003eAccording to Mendelian Randomization analysis, FIBCD1 was identified as a potential risk gene associated with the prognosis of breast cancer (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eA-B). Kaplan\u0026ndash;Meier survival curves further confirmed that low FIBCD1 expression was correlated with longer OS (Log-rank p\u0026thinsp;=\u0026thinsp;0.03, Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eC). Furthermore, ELISA results demonstrated that the expression level of FIBCD1 protein in the peripheral blood of patients with a ALN-high burden was elevated (p\u0026thinsp;\u0026lt;\u0026thinsp;0.01, Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eD), confirming that FIBCD1 is also a potential risk gene for ALN metastasis.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eAssociations between Proteomics and FIBCD1\u003c/b\u003e\u003c/p\u003e\u003cp\u003eFrom the peripheral blood Olink Inflammation panels data, 75 immune-related features were extracted. Pearson correlation analysis showed that FIBCD1 expression was positively correlated with OSM (cor\u0026thinsp;=\u0026thinsp;0.69, p\u0026thinsp;=\u0026thinsp;1.87E-12), TNFSF14 (cor\u0026thinsp;=\u0026thinsp;0.62, p\u0026thinsp;=\u0026thinsp;6.61E-10), CASP-8 (cor\u0026thinsp;=\u0026thinsp;0.61, p\u0026thinsp;=\u0026thinsp;1.81E-09), and TGF-alpha (cor\u0026thinsp;=\u0026thinsp;0.59, p\u0026thinsp;=\u0026thinsp;4.22E-09), but negatively correlated with CCL11 (cor\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.40, p\u0026thinsp;=\u0026thinsp;0.0002) and MCP-4 (cor\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.34, p\u0026thinsp;=\u0026thinsp;0.0002) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eE).\u003c/p\u003e\u003cp\u003e\u003cb\u003eAssociations between Radiomics and FIBCD1\u003c/b\u003e\u003c/p\u003e\u003cp\u003eFrom each volume of interest (VOI) in MRI, 944 radiomic features were extracted. Spearman correlation analysis revealed that FIBCD1 expression was positively associated with original_shape_Sphericity (Cor\u0026thinsp;=\u0026thinsp;0.31, p\u0026thinsp;\u0026lt;\u0026thinsp;0.05), a metric quantifying the roundness of the tumor shape, and negatively correlated with wavelet.LLH_glcm_Idn (Cor = -0.27, p\u0026thinsp;\u0026lt;\u0026thinsp;0.05), which measures the local homogeneity of gray-level intensities within the tumor texture. These radiomic features notably overlap with the key predictors previously utilized in our MRI-based model for ALN burden (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eF).\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this study, we developed a macro\u0026ndash;micro\u0026ndash;macro radiogenomic framework that links MRI-derived imaging phenotypes to immune-modulating molecular mechanisms in breast cancer. Through radiomics-guided gene screening and multi-omics validation, we identified FIBCD1 as a central prognostic biomarker associated with immune suppression and adverse clinical outcomes. Notably, this study ultimately identified FIBCD1 as a hub gene highly associated with poor prognosis in breast cancer and potentially a key biological marker influencing tumor imaging phenotypes.\u003c/p\u003e\u003cp\u003eUnder the guidance of radiomics, the performance of our prognostic model is comparable to other studies[11, 19, 20]. However, unlike previous research, our study provides biological explanations for the relationship between prognosis and underlying molecular mechanisms[11, 19, 20]. We found that patients in the high-risk group often exhibited typical features of chronic inflammation activation and immune evasion, with significant upregulation of inflammatory cytokines such as IL-6, indicating weakened immune responses and reduced anti-tumor capabilities[22, 23]. In contrast, the low-risk group demonstrated an IFN-γ-dominated immune profile with enhanced T-cell activity, which aligns with better prognoses[24, 25]. Drug sensitivity analysis further revealed that high-risk patients exhibited potential resistance to targeted drugs such as Palbociclib, Ribociclib, and Niraparib, suggesting that alternative therapeutic strategies should be designed for this subgroup[26]. Additionally, GSEA enrichment analysis showed that immune pathways were significantly enriched in the low-risk group, whereas metabolic-related pathways were upregulated in the high-risk group, suggesting that metabolic reprogramming may be another potential mechanism influencing prognosis[27]. From a clinical perspective, combined immune modulation and metabolic-targeted strategies may provide a more effective treatment approach for high-risk patients[28].\u003c/p\u003e\u003cp\u003eAt the micro-level, we identified FIBCD1 as a hub driver of high-risk scores. The prognostic value of this gene is supported by multiple lines of evidence: First, Mendelian randomization analysis demonstrated a causal relationship between FIBCD1 and breast cancer prognosis, suggesting that its expression is genetically regulated and directly impacts survival outcomes[29]. Second, ELISA experiments validated that FIBCD1 levels were significantly elevated in the peripheral blood of patients with high axillary lymph node (ALN) burden. Third, Olink proteomics analysis showed that high FIBCD1 expression was positively correlated with various pro-inflammatory and immune-suppressive proteins (e.g., TGF-α, CASP-8, EN-RAGE), while it was negatively correlated with chemokines (e.g., CCL11, MCP-4). These findings suggest that FIBCD1 may contribute to the establishment of an \"immune cold tumor\" phenotype by inhibiting immune cell recruitment and weakening immune responses, thereby promoting immune evasion and tumor progression[30].\u003c/p\u003e\u003cp\u003eOf greater clinical translational significance, we achieved a closed-loop analysis from molecular mechanisms to imaging features at the macro level. Imaging-molecular correlation analysis revealed that FIBCD1 expression was positively correlated with tumor sphericity (original_shape_Sphericity) and negatively correlated with texture signal homogeneity (wavelet.LLH_glcm_Idn). Specifically, high-expression patients tended to exhibit mass-like tumors with higher sphericity and greater textural heterogeneity, which are the hub radiomic features we identified in the ALN prediction model[15]. We further validated this imaging-molecular-clinical consistency through two representative cases (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e), demonstrating that FIBCD1 is not only a potential molecular biomarker but also likely plays a role in regulating tumor morphology, thus influencing the formation of imaging features and the accuracy of prognostic predictions[10].\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eThe principal innovation of this study is the development of a macro\u0026ndash;micro\u0026ndash;macro closed-loop framework that integrates imaging feature\u0026ndash;guided gene screening with molecular validation, thereby enhancing the biological interpretability and clinical applicability of breast cancer prognostic assessment. Compared to prior work that relies solely on either imaging or gene expression features, our radiomic-guided gene selection of candidate genes and in-depth immune-microenvironment analysis may prove especially advantageous[31\u0026ndash;33]. Moreover, the introduction of Mendelian randomization helps mitigate confounding factors, thus adding further weight to the causal association between FIBCD1 and breast cancer prognosis[34].\u003c/p\u003e\u003cp\u003eThis study has certain limitations. First, our study primarily utilized publicly available datasets, which may introduce selection bias and limit the generalizability of the results. Second, while we identified associations between FIBCD1 expression and imaging features, the underlying biological mechanisms remain to be elucidated through functional experiments. Third, the study's retrospective nature necessitates prospective validation in independent cohorts to confirm the clinical utility of the prognostic model.\u003c/p\u003e\u003cp\u003eIn summary, we developed a precision prognostic model for breast cancer through a closed-loop macro\u0026ndash;micro\u0026ndash;macro framework, integrating radiomics-guided gene selection, transcriptomic data, and biological validation. This model elucidates FIBCD1\u0026rsquo;s role in tumor progression and highlights the influence of the immune microenvironment on patient outcomes. Notably, the significant correlation between FIBCD1 expression and key radiomic features provides novel evidence linking imaging phenotypes to underlying molecular biology. Our findings not only offer new perspectives for individualized treatment strategies but also establish a foundation for further exploration of tumor metastasis mechanisms and immunotherapeutic targets.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eALN Axillary Lymph Node\u003c/p\u003e\n\u003cp\u003eAUC Area Under the Curve\u003c/p\u003e\n\u003cp\u003eC-index Concordance Index\u003c/p\u003e\n\u003cp\u003eDEG Differentially Expressed Gene\u003c/p\u003e\n\u003cp\u003eDCE-MRI Dynamic Contrast-Enhanced Magnetic Resonance Imaging\u003c/p\u003e\n\u003cp\u003eFDR False Discovery Rate\u003c/p\u003e\n\u003cp\u003eGSEA Gene Set Enrichment Analysis\u003c/p\u003e\n\u003cp\u003eGWAS Genome-Wide Association Study\u003c/p\u003e\n\u003cp\u003eHEIDI Heterogeneity in Dependent Instruments\u003c/p\u003e\n\u003cp\u003eIC50 Half-Maximal Inhibitory Concentration\u003c/p\u003e\n\u003cp\u003eICC Intraclass Correlation Coefficient\u003c/p\u003e\n\u003cp\u003eNPX Normalized Protein Expression\u003c/p\u003e\n\u003cp\u003eOS Overall Survival\u003c/p\u003e\n\u003cp\u003eOSM Oncostatin M\u003c/p\u003e\n\u003cp\u003ePEA Proximity Extension Assay\u003c/p\u003e\n\u003cp\u003eROC Receiver Operating Characteristic\u003c/p\u003e\n\u003cp\u003eSMR Summary-data-based Mendelian Randomization\u003c/p\u003e\n\u003cp\u003eTIDE Tumor Immune Dysfunction and Exclusion\u003c/p\u003e\n\u003cp\u003eTME Tumor Microenvironment\u003c/p\u003e\n\u003cp\u003eVOI Volume of Interest\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthical Approval and Consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was approved by the Institutional Review Board of Jiaxing TCM Hospital Affiliated to Zhejiang Chinese Medical University (IRB No: SL-2022-0369, Approval Date: October 18, 2022). All participants provided written informed consent prior to inclusion in the study.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll authors have read and approved the final manuscript and consent to its publication.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors gratefully acknowledge the support of the Guangdong Provincial Key Laboratory of Artificial Intelligence in Medical Image Analysis and Application for providing technical guidance and infrastructure during the course of this study.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026apos; contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eM.H. , K.L., and H.S. conceived and designed the study; K.L., Z.M., X.H. and X.C. acquired the data; W.J. and J.H. did the statistical analyses; Y.Y. and X.H. implemented quality control of data and the algorithms; All authors analyzed and interpreted the data; K.L., Z.M., X.C. and H.S.. prepared the first draft of the manuscript; M.H. revised the manuscript; All authors contributed to manuscript preparation.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe research was supported by the grants of Zhejiang Provincial Natural Science Foundation of China (Grant No. LQN25H180008), Medical Science and Technology Project of Zhejiang Province (Grant No. 2025KY366, 2024KY454, 2025KY1626), Public Welfare Research Project of Jiaxing (Grant No. 2024AY10026), Zhejiang Traditional Chinese Medicine Administration (Grant No. 2024ZL1058) and Shanghai Municipal Health Commission Health Industry Clinical Research Special Program (Youth Project) (20244Y0003).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets generated and/or analyzed during the current study have been registered with the Chinese Clinical Trial Registry (ChiCTR2500103228). While the data are not publicly available at the time of submission, they will be uploaded to the ChiCTR platform after project completion. During the interim period, data are available from the corresponding author upon reasonable request. Additionally, the code used in the analyses is available on GitHub at https://github.com/idcast-Tracy/Image-genomics.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eSiegel RL, Kratzer TB, Giaquinto AN, Sung H, Jemal A. Cancer statistics, 2025. CA Cancer J Clin. 2025;75:10\u0026ndash;45.\u003c/li\u003e\n\u003cli\u003eWang G, Wang S, Song W, Lu C, Chen Z, He L, et al. Integrating multi-omics data reveals the antitumor role and clinical benefits of gamma-delta T cells in triple-negative breast cancer. BMC Cancer. 2025;25:623.\u003c/li\u003e\n\u003cli\u003eMeattini I, Becherini C, Caini S, Coles CE, Cortes J, Curigliano G, et al. International multidisciplinary consensus on the integration of radiotherapy with new systemic treatments for breast cancer: European Society for Radiotherapy and Oncology (ESTRO)-endorsed recommendations. Lancet Oncol. 2024;25:e73\u0026ndash;83.\u003c/li\u003e\n\u003cli\u003eRayson VC, Harris MA, Savas P, Hun ML, Virassamy B, Salgado R, et al. The anti-cancer immune response in breast cancer: current and emerging biomarkers and treatments. Trends Cancer. 2024;10:490\u0026ndash;506.\u003c/li\u003e\n\u003cli\u003eTeng J, Zhang H, Liu W, Shu X-O, Ye F. A Dynamic Bayesian Model for Breast Cancer Survival Prediction. IEEE J Biomed Health Inform. 2022;26:5716\u0026ndash;27.\u003c/li\u003e\n\u003cli\u003eWang R, Zhu Y, Liu X, Liao X, He J, Niu L. The Clinicopathological features and survival outcomes of patients with different metastatic sites in stage IV breast cancer. BMC Cancer. 2019;19:1091.\u003c/li\u003e\n\u003cli\u003eFan R, Chen Y, Nechuta S, Cai H, Gu K, Shi L, et al. Prediction models for breast cancer prognosis among Asian women. Cancer. 2021;127:1758\u0026ndash;69.\u003c/li\u003e\n\u003cli\u003eJiang Y-Z, Ma D, Jin X, Xiao Y, Yu Y, Shi J, et al. Integrated multiomic profiling of breast cancer in the Chinese population reveals patient stratification and therapeutic vulnerabilities. Nat Cancer. 2024;5:673\u0026ndash;90.\u003c/li\u003e\n\u003cli\u003eSammut S-J, Crispin-Ortuzar M, Chin S-F, Provenzano E, Bardwell HA, Ma W, et al. Multi-omic machine learning predictor of breast cancer therapy response. Nature. 2022;601:623\u0026ndash;9.\u003c/li\u003e\n\u003cli\u003eQi Y-J, Su G-H, You C, Zhang X, Xiao Y, Jiang Y-Z, et al. Radiomics in breast cancer: Current advances and future directions. Cell Rep Med. 2024;5:101719.\u003c/li\u003e\n\u003cli\u003eWu X, Lu X, Zhang W, Zhong X, Bu H, Zhang Z. Development and validation of a 10-gene signature for predicting recurrence risk in HR+/HER2- early breast cancer undergoing chemo-endocrine therapy. Breast. 2025;82:104484.\u003c/li\u003e\n\u003cli\u003eVerma VK, Beevi SS, Nair RA, Kumar A, Kiran R, Alexander LE, et al. MicroRNA signatures differentiate types, grades, and stages of breast invasive ductal carcinoma (IDC): miRNA-target interacting signaling pathways. Cell Commun Signal. 2024;22:100.\u003c/li\u003e\n\u003cli\u003eChen Z, Yang C, Wang J, Wang B, He L, Liu Z, et al. Integrating of Radiomics and Single-cell Transcriptomics Uncovers the Crucial Role of \u0026gamma;\u0026delta; T Cells in Lung Adenocarcinoma. J Cancer. 2025;16:1167\u0026ndash;80.\u003c/li\u003e\n\u003cli\u003eCorti A, Lenoci D, Corino VDA, Mattavelli D, Ravanelli M, Poli T, et al. Interplay between MRI radiomics and immune gene expression signatures in oral squamous cell carcinoma. Sci Rep. 2025;15:12622.\u003c/li\u003e\n\u003cli\u003eHong M, Fan S, Xu Z, Fang Z, Ling K, Lai P, et al. MRI radiomics and biological correlations for predicting axillary lymph node burden in early-stage breast cancer. J Transl Med. 2024;22:826.\u003c/li\u003e\n\u003cli\u003eLi Z, Gao J, Zhou H, Li X, Zheng T, Lin F, et al. Multiregional dynamic contrast-enhanced MRI-based integrated system for predicting pathological complete response of axillary lymph node to neoadjuvant chemotherapy in breast cancer: multicentre study. EBioMedicine. 2024;107:105311.\u003c/li\u003e\n\u003cli\u003eKimbung S, Loman N, Hedenfalk I. Clinical and molecular complexity of breast cancer metastases. Semin Cancer Biol. 2015;35:85\u0026ndash;95.\u003c/li\u003e\n\u003cli\u003eSchnitt SJ. Classification and prognosis of invasive breast cancer: from morphology to molecular taxonomy. Mod Pathol. 2010;23 Suppl 2:S60-64.\u003c/li\u003e\n\u003cli\u003eGuan J, Fan M, Li L. MVNMF: Multiview nonnegative matrix factorization for radio-multigenomic analysis in breast cancer prognosis. Med Image Anal. 2025;103:103566.\u003c/li\u003e\n\u003cli\u003eWu X, Li Y, Chen J, Chen J, Zhang W, Lu X, et al. Multimodal recurrence risk prediction model for HR+/HER2- early breast cancer following adjuvant chemo-endocrine therapy: integrating pathology image and clinicalpathological features. Breast Cancer Res. 2025;27:27.\u003c/li\u003e\n\u003cli\u003eThorsson V, Gibbs DL, Brown SD, Wolf D, Bortone DS, Ou Yang T-H, et al. The Immune Landscape of Cancer. Immunity. 2018;48:812-830.e14.\u003c/li\u003e\n\u003cli\u003eManore SG, Doheny DL, Wong GL, Lo H-W. IL-6/JAK/STAT3 Signaling in Breast Cancer Metastasis: Biology and Treatment. Front Oncol. 2022;12:866014.\u003c/li\u003e\n\u003cli\u003eWeng Y-S, Tseng H-Y, Chen Y-A, Shen P-C, Al Haq AT, Chen L-M, et al. MCT-1/miR-34a/IL-6/IL-6R signaling axis promotes EMT progression, cancer stemness and M2 macrophage polarization in triple-negative breast cancer. Mol Cancer. 2019;18:42.\u003c/li\u003e\n\u003cli\u003eCai J, Zhong M, Xu J, Cheng H, Xu S. Codelivery of triptolide and IFN-\u0026gamma; to boost antitumor immunity for triple-negative breast cancer. Int Immunopharmacol. 2023;120:110346.\u003c/li\u003e\n\u003cli\u003eCastro F, Pinto ML, Leite Pereira C, Serre K, Costa \u0026Acirc;M, Cavadas B, et al. Chitosan/\u0026gamma;-PGA nanoparticles and IFN-\u0026gamma; immunotherapy: A dual approach for triple-negative breast cancer treatment. J Control Release. 2025;379:621\u0026ndash;35.\u003c/li\u003e\n\u003cli\u003eLi S, Zhang Y, Wang N, Guo R, Liu Q, Lv C, et al. Pan-cancer analysis reveals synergistic effects of CDK4/6i and PARPi combination treatment in RB-proficient and RB-deficient breast cancer cells. Cell Death Dis. 2020;11:219.\u003c/li\u003e\n\u003cli\u003eRen W, Yu Y, Wang T, Wang X, Su K, Wang Y, et al. Comprehensive analysis of metabolism-related gene biomarkers reveals their impact on the diagnosis and prognosis of triple-negative breast cancer. BMC Cancer. 2025;25:668.\u003c/li\u003e\n\u003cli\u003eXia L, Oyang L, Lin J, Tan S, Han Y, Wu N, et al. The cancer metabolic reprogramming and immune response. Mol Cancer. 2021;20:28.\u003c/li\u003e\n\u003cli\u003eEem M, S le C. Instrumental Variables Analysis and Mendelian Randomization for Causal Inference. The Journal of infectious diseases. 2025;231.\u003c/li\u003e\n\u003cli\u003eGalon J, Bruni D. Approaches to treat immune hot, altered and cold tumours with combination immunotherapies. Nat Rev Drug Discov. 2019;18:197\u0026ndash;218.\u003c/li\u003e\n\u003cli\u003eTao Y, Wang Q, Guo S, Liu J, Cao Y. m6A related metabolic genes in breast cancer and their relationship with prognosis. Int Immunopharmacol. 2025;148:114121.\u003c/li\u003e\n\u003cli\u003eHuang R, Li Y, Lin K, Zheng L, Zhu X, Huang L, et al. A novel glycolysis-related gene signature for predicting prognosis and immunotherapy efficacy in breast cancer. Front Immunol. 2025;16:1512859.\u003c/li\u003e\n\u003cli\u003eWenwen null, Jiang Z, Liu J, Liu D, Li Y, He Y, et al. Integrating ultrasound radiomics and clinicopathological features for machine learning-based survival prediction in patients with nonmetastatic triple-negative breast cancer. BMC Cancer. 2025;25:291.\u003c/li\u003e\n\u003cli\u003eEvans DM, Davey Smith G. Mendelian Randomization: New Applications in the Coming Age of Hypothesis-Free Causality. Annu Rev Genomics Hum Genet. 2015;16:327\u0026ndash;50.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":true,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"journal-of-translational-medicine","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"jtrm","sideBox":"Learn more about [Journal of Translational Medicine](http://translational-medicine.biomedcentral.com)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/jtrm/default.aspx","title":"Journal of Translational Medicine","twitterHandle":"@BioMedCentral","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Breast Cancer, Radiomics, Prognostic Model, Transcriptomics, FIBCD1, Immune Microenvironment","lastPublishedDoi":"10.21203/rs.3.rs-7024804/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7024804/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cb\u003eBackground\u003c/b\u003e\u003c/p\u003e\u003cp\u003eBreast cancer prognosis remains challenging due to tumor heterogeneity and the limited predictive power of conventional clinical models. Integrating imaging features with molecular data may improve individualized risk stratification and clinical decision-making.\u003c/p\u003e\u003cp\u003e\u003cb\u003eMethods\u003c/b\u003e\u003c/p\u003e\u003cp\u003eWe developed a closed-loop prognostic model based on a macro\u0026ndash;micro\u0026ndash;macro radiogenomic framework that combines MRI-based radiomics with transcriptomic and proteomic data. A total of 788 radiomics-guided candidate genes were screened. Prognostic gene signatures were identified using multiple machine learning algorithms and validated in TCGA and GEO cohorts. We further analyzed immune infiltration, drug sensitivity, and gene enrichment profiles across risk groups. Causal relationships between gene expression and survival were assessed using Mendelian randomization. FIBCD1 expression was validated in patient plasma using ELISA, and Olink proteomics and radiomic analyses were conducted for biological interpretation.\u003c/p\u003e\u003cp\u003e\u003cb\u003eResults\u003c/b\u003e\u003c/p\u003e\u003cp\u003eWe identified a 10-gene prognostic signature. The combined Elastic Net and stepwise Cox regression model achieved the highest concordance index (C-index\u0026thinsp;=\u0026thinsp;0.645). High-risk patients showed reduced immune activation, increased expression of pro-inflammatory cytokines such as IL-6, and shorter survival. FIBCD1 emerged as a key prognostic gene with a significant causal association with survival. Its expression was significantly elevated in high-risk plasma samples, positively correlated with inflammatory proteins (e.g., OSM, CASP-8), and associated with specific radiomic features such as tumor sphericity.\u003c/p\u003e\u003cp\u003e\u003cb\u003eConclusions\u003c/b\u003e\u003c/p\u003e\u003cp\u003eOur findings establish a novel radiogenomic strategy that bridges MRI-derived imaging phenotypes with molecular mechanisms. FIBCD1 may serve as an immune-modulating prognostic biomarker linked to imaging characteristics, providing new insights into non-invasive breast cancer risk assessment and therapeutic targeting.\u003c/p\u003e","manuscriptTitle":"A Macro–micro–macro Radiogenomic Framework Identifies FIBCD1 as a Key Immune-modulating Biomarker in Breast Cancer","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-08-07 10:02:23","doi":"10.21203/rs.3.rs-7024804/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewerAgreed","content":"","date":"2025-08-03T16:56:05+00:00","index":0,"fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-08-03T16:41:10+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-07-04T15:49:44+00:00","index":"","fulltext":""},{"type":"submitted","content":"Journal of Translational Medicine","date":"2025-07-01T23:45:34+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"journal-of-translational-medicine","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"jtrm","sideBox":"Learn more about [Journal of Translational Medicine](http://translational-medicine.biomedcentral.com)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/jtrm/default.aspx","title":"Journal of Translational Medicine","twitterHandle":"@BioMedCentral","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"ef06ae4d-69a8-46ba-b139-1657bc734633","owner":[],"postedDate":"August 7th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2025-12-01T16:03:47+00:00","versionOfRecord":{"articleIdentity":"rs-7024804","link":"https://doi.org/10.1186/s12967-025-07389-z","journal":{"identity":"journal-of-translational-medicine","isVorOnly":false,"title":"Journal of Translational Medicine"},"publishedOn":"2025-11-25 15:58:19","publishedOnDateReadable":"November 25th, 2025"},"versionCreatedAt":"2025-08-07 10:02:23","video":"","vorDoi":"10.1186/s12967-025-07389-z","vorDoiUrl":"https://doi.org/10.1186/s12967-025-07389-z","workflowStages":[]},"version":"v1","identity":"rs-7024804","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7024804","identity":"rs-7024804","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}