A novel ferroptosis-related gene signature associated with cuproptosis for predicting overall survival in breast cancer patients

Purpose Ferroptosis and cuproptosis are both metal-dependent regulated cell death that play an important role in cancer. However, the expression patterns and the prognostic values of ferroptosis-related genes (FRGs) associated with cuproptosis in breast cancer (BC) are largely unknown. This study aims to explore the prognostic value of cuproptosis-related FRGs and their relationship with tumor microenvironments in BC. Methods The clinical and RNA sequencing data of BC patients from TCGA, META-BRIC and GEO databases were analyzed. The least absolute shrinkage and selection operator regression analysis was used to establish prognostic signatures based on cu-proptosis-related FRGs. The overall survival between risk subgroups was assessed by Kaplan-Meier analysis. The changes in risk score during neoadjuvant chemotherapy, and differences in immune cells, immune checkpoints, and drug sensitivity between risk subgroups were also analyzed in this study. Results A successful development of a prognostic signature based on cuproptosis-related FRGs in the TCGA cohort was achieved and it was validated in the METABRIC cohort. Gene set enrichment analysis results revealed the enrichment of steroid biosynthesis and ABC transporters in the high-risk group. Moreover, the signature was also found to be associated with immune cells and immune checkpoints. Lower risk score in patients after neoadjuvant chemotherapy and higher sensitivity of the high-risk group to AKT inhibitor VIII and cisplatin was also observed. Conclusion Cuproptosis-related FRGs can be used as a novel prognostic signature for predicting the overall survival of BC patients. This can provide meaningful insights into the selection of immuno-therapy and antitumor drugs for BC.


INTRODUCTION
Breast cancer (BC) is one of the most frequently diagnosed cancer in women worldwide, with an extremely high rate of mortality (Azamjah et al., 2019).Its incidence and death rates have increased over recent years making it the second most common cause of mortality in women (DeSantis et al., 2019).Its progressive impact on younger individuals is a cause of concern.Although the etiology and mechanism of BC are not completely understood, biological features including cell infiltration and genetic abnormalities are thought to be the major causes associated with the progression and metastasis of the disease (Feng et al., 2018;Loibl et al., 2021).Although there have been advancements in diagnosis, chemotherapy, endocrine and targeted therapies, the median survival in the case of advanced BC is only 31.8 months (Caswell-Jin et al., 2018;Sung et al., 2021).Early diagnosis and intervention are of paramount importance.Therefore, research on novel prognostic markers and therapeutic targets will provide improved opportunities for individualized treatment of BC.
Metals are essential components of metabolic processes.A significant portion of the proteases requires binding to metals, such as calcium, magnesium, iron, and copper for proper functioning (Waldron et al., 2009).However, dysregulation of metal metabolism can lead to cell death.Ferroptosis, unlike apoptosis and autophagy, is an iron-dependent process of regulated cell death that is characterized by intracellular buildup of reactive oxygen species (ROS) and products of lipid peroxidation (Huang et al., 2020).The primary mechanism of ferroptosis is the catalysis of the highly expressed unsaturated fatty acids on the cell membrane to produce liposome peroxidation, resulting in cell death, under the influence of divalent iron or ester oxygenase (Dixon et al., 2012).Ferroptosis-related genes (FRGs) play an important role in the development of various cancers, including lung cancer, melanoma, renal cell carcinoma, BC, etc. (Friedmann Angeli et al., 2019;Wang et al., 2020;Zhao et al., 2022).Activation of ferroptosis has been shown to prevent the growth and proliferation of tumors (Hassannia et al., 2019).Therefore, ferroptosis can be a potential target for cancer therapy, especially in patients who have grown resistant to conventional forms of therapy (Hassannia et al., 2019;Zhang et al., 2022).For example, the sensitivity of triple-negative BC cells to gefitinib was enhanced by the inhibition of GPX4 activation of Ferroptosis (Song et al., 2020).Cuproptosis is a method of copper-induced regulated cell death.The direct interaction between copper ions and lipidated protein compo- nents in the tricarboxylic acid cycle interferes with the iron-sulfur cluster proteins in the respiratory chain complex which ultimately leads to proteotoxic stress and cell death (Kahlson & Dixon, 2022).This could be another regulatory mechanism for the development of cancer (Jiang et al., 2022).Iron-sulfur cluster proteins are able to maintain iron homeostasis in mitochondria (Pain & Dancis, 2016).Several mitochondrial proteins (including NFS1, ISCU, CISD1, and CISD2) require iron for iron-sulfur cluster biogenesis reactions for the negative regulation of ferroptosis (Yuan et al., 2016;Alvarez et al., 2017;Kim et al., 2018;Du et al., 2019).Moreover, elesclomol stimulates copper retention and the subsequent buildup of ROS via the degradation of ATP7A, which promotes ferroptosis in colorectal cancer cells (Gao et al., 2021).Cuprizone, which is a copper chelator, regulates demyelination by the stimulation of ferroptosis and eventually leads to the loss of oligodendrocytes (Jhelum et al., 2020).Therefore, an association between iron and copper metabolism has been well-established in scientific literature.These findings highlight the role of ferroptosis and cuproptosis as potential targets for the treatment of cancer.
In this study, a predictive model based on cuproptosis-related FRGs was built successfully to evaluate the prognosis of patients with BC and the predictive performance of the model was further validated.The findings of this study may enhance the effectiveness of individualized treatment and prognostic evaluations in patients with BC.

Patient and gene set data collection
Clinical and RNA-sequencing data of 1097 BC patients were obtained from the cancer genome atlas (TCGA) database and 1053 patients containing complete clinical data including overall survival, age, gender, T stage, N stage, M stage, and TNM stage, were obtained after screening, while 44 patients who lacked complete clinical data were excluded.The intrinsic subtype was defined by PAM 50 (Parker et al., 2009).Moreover, 1904 cases of RNA-sequencing and corresponding clinical data were obtained from the molecular taxonomy of breast cancer international consortium (METABRIC) database as the validation set for prognosis, and datasets GSE18728 and GSE87455 from gene expression omnibus (GEO) database were used to compare changes in risk score before and after neoadjuvant chemotherapy.Furthermore, 13 cuproptosis-related genes were obtained from the cuproptosis-related study (Tsvetkov et al., 2022), and 259 FRGs were downloaded from the FerrDb website (Zhou & Bao, 2020).

Identification and prognosis of cuproptosis-related FRGs
The correlation between the expression of cuproptosis-related genes and FRGs was determined using the Pearson correlation coefficients.The criteria used to identify cuproptosis-related FRGs was of P value <0.001 and the absolute value of the Pearson correlation coefficient >0.3(|R|>0.3).The univariate Cox regression analysis was used to screen the prognosis-related genes.

Building and validation of a prognostic model
Based on the outcome of univariate Cox regression analysis, the least absolute shrinkage and selection opera-tor (LASSO) regression method was used for the identification of the best survival-related genes via glmnet R package (Tibshirani, 1997;Wang & Liu, 2020).The prognostic risk score formula was created using the coefficients obtained from LASSO regression and the expression levels of genes.The formula can be mathematically represented as: Risk Score = ∑ corresponding regression coefficient * expression of the gene.The risk scores of the patients with BC were determined and the patients were grouped into two groups: the high-risk group and the low-risk group using the median risk score as the cutoff.The Kaplan-Meier survival analysis was used to demonstrate the presence of survival differences between the two groups.A receiver operator characteristic (ROC) curve via survivalROC R package was constructed to evaluate the effectiveness of the prognostic model and the METABRIC cohort was used for further validation.

Establishment of a nomogram based on risk score and clinical characteristics
Univariate and multivariate COX regression analyses were used for the evaluation of risk scores and clinical characteristics as independent prognostic factors.A nomogram based on risk score and clinical characteristics was established to predict the probability of 3-, 5and 10-year overall survival (OS) for patients with BC, and the performance of the nomogram was assessed by calibration curves.

Functional enrichment analysis, protein-protein interaction (PPI), and gene set enrichment analysis (GSEA)
The Gene Ontology (GO) analysis was carried out for the identification of biological activities, molecular mechanisms, and cellular components via clusterprofiler R package.Additionally, the signaling pathways were observed using the Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis.Cuproptosis-related FRGs were submitted to the STRING database (http://www.stringdb.org/)for obtaining insights into PPI.The Cytoscape program was used to build and render PPI networks.Using the MCODE plug-in, the most important modules were selected based on MCODE score >5.GSEA was carried out to investigate potential enrichment pathways in the high-and low-risk groups.P value <0.05 was considered statistically significant.

Immune cell analysis
Different algorithms including CIBERSORT (Newman et al., 2015), CIBERSORT-ABS, EPIC (Racle et al., 2017), MCP-counter (Becht et al., 2016), QUAN-TISEQ (Finotello et al., 2019), TIMER (Li et al., 2017), and XCELL (Aran et al., 2017) were compared to assess the level of immune cell infiltration between the highand low-risk groups.The correlation between immune cells and cuproptosis-related FRGs was identified by using the TIMER database.Potential immune checkpoints were obtained from scientific literature and differences in expression in immune checkpoints among high-and low-risk groups were explored using the ggpubr and limma packages.

Drug sensitivity analysis
To compare the difference in drug sensitivity between the two groups, the half-maximal inhibitory concentration (IC50) of drugs was assessed by the pRRophetic Ferroptosis-related gene signature associated with cuproptosis package.P value <0.05 was considered statistically significant.

Statistics analysis
Statistical analysis was performed using R version 4.2.1.The Wilcoxon test was used to compare expression differences between the two groups.P value <0.05 was considered to be a statistically significant difference.

Identification of cuproptosis-related FRGs and their prognostic value in the TCGA cohort
The design of the study has been depicted using the flowchart (Fig. 1).A total of 125 FRGs were confirmed to be correlated with cuproptosis-related genes in the TCGA dataset (Supplementary Fig. S1 at https://ojs.ptbioch.edu.pl/index.php/abp/).Univariate Cox regression analysis was used to explore the prognostic value of these cuproptosis-related FRGs, and the results revealed that 15 genes had prognostic value (Fig. 2).

Establishment and validation of cuproptosis-related FRG signature
Based on the 15 cuproptosis-related FRGs of prognostic value described above, LASSO Cox regression analysis was performed at the minimum λ value to create a risk model consisting of 11 genes (Supplementary Fig. Patients with BC were divided into the high-risk group (n=526) and the low-risk group (n=527) based on the median risk score.Overall survival was significantly shorter in the high-risk group than in the low-risk group (P<0.001),suggesting a negative correlation between risk score and prognosis (Fig. 3A).Time-dependent ROC curve analysis verified the accuracy of the prognostic signature of patients with BC, with AUCs reaching 0.669 for 3 years, 0.643 for 5 years, and 0.711 for 10 years (Fig. 3B).
To verify the stability of the cuproptosis-related FRG signature, risk scores were calculated for patients in the METABRIC cohort by using the risk score formula ob-   tained from the TCGA cohort.K-M survival analysis revealed a poor prognosis in the high-risk group (Fig. 4A, P<0.001).The result obtained was consistent with the TCGA dataset.The time-dependent ROC curves showed AUCs of 0.587 at 3 years, 0.596 at 5 years, and 0.559 at 10 years (Fig. 4B).

The signature based on cuproptosis-related FRGs was an independent indicator of BC prognosis
For further assessment of the independent prognostic value of cuproptosis-related FRG signature, univariate and multivariate Cox regression analyses of clinical characteristics, including age, gender, and TNM stage were performed.Univariate analysis revealed that higher risk scores, T stage, N stage, and M stage were significantly associated with adverse OS in patients with BC (Fig. 5A), and negative estrogen receptor (ER) and progesterone receptor (PR) were suggestive of poor prognosis.The risk score, age, N stage, and M stage were shown to be independent risk factors for OS in the multivariate Cox analysis (Fig. 5B).Therefore, the cuproptosis-related FRG signature was an independent prognostic indicator for patients with BC.

Association between the risk score and clinical characteristics
For the assessment of the impact of cuproptosis-related FRG signature in the development and progression of BC, the association between risk score and clinical characteristics was explored.Chi-square test analysis revealed significant differences between risk groups in T stage, N stage, TNM stage, PR, ER, human epidermal growth factor receptor (HER-2), and intrinsic subtype (Fig. 6).In addition, HER-2-enriched, ER-negative, PR-negative, HER-2-positive, higher N stage, and TNM stage indicated a higher risk score (Fig. 7).Stratified analysis was carried out for the identification of the prognostic value of the cuproptosis-related FRG signature in subgroups.The results revealed that this signature had significant prognostic efficacy in age <=65 (P=0.001),age >65 (P=0.009),female (P<0.001),T1 stage (P=0.011),T2-T4 stage (P=0.001),N0 stage (P<0.001),N1-N3 stage (P=0.018),M0 stage (P<0.001),stage I (P=0.047) and II-IV (P<0.001).However, no significant prognostic value was observed in the male patients and the M1 stages (P>0.05)(Fig. 8).

Establishment of nomogram
A nomogram was constructed using risk score and clinical characteristics, including age, T stage, N stage, and M stage (Fig. 9A).The nomogram predicted the survival of BC patients at 3, 5, and 10 years.The calibration curves further validated the consistency of the actual OS of the patients with the predictions of the nomogram (Fig. 9B).

Functional enrichment and PPI of cuproptosis-related FRGs
The potential biological functions involved in 125 cuproptosis-related FRGs were analyzed using the GO and KEGG databases.GO analysis revealed the association of FRGs with biological processes like response to oxidative stress, cellular response to chemical stress, and response to nutrient levels (Supplementary Fig. S3A at https://ojs.ptbioch.edu.pl/index.php/abp/).In addition, in the KEGG pathway, the results revealed the involvement of FRGs in lipid and atherosclerosis, autophagy, FoxO signaling pathway, central carbon metabolism in cancer, and ferroptosis (Supplementary Fig. S3B at https://ojs.ptbioch.edu.pl/index.php/abp/).STRING database revealed that the PPI network of cuproptosisrelated FRGs consisted of 125 nodes and 276 edges (Supplementary Fig. S3C at https://ojs.ptbioch.edu.pl/index.php/abp/).The most significant module consisted of 11 cuproptosis-related FRGs, including 11 nodes and 66 edges.

Immune infiltration level of cuproptosis-related FRG signature
The heatmap depicted the relationship between the cuproptosis-related FRG signature and immune cell subgroups according to the analysis of CIBERSORT, CIBERSORT-ABS, EPIC, MCP-counter, QUANTISEQ, TIMER, and XCELL algorithms (Fig. 10).The CIB-ERSORT result showed a higher proportion of CD4+ memory T cells, resting NK cells, M0 macrophages, M1 macrophages, M2 macrophages, eosinophils, and neutrophils in the high-risk group, whereas the low-risk group had higher proportions of memory B cells, CD8+ T cells, T follicular helper cells, Tregs, activated NK cells, monocytes, myeloid dendritic cells, activated mast cells, and resting mast cells (Supplementary Fig. S4 at https:// ojs.ptbioch.edu.pl/index.php/abp/).In addition, the TIMER database was used to analyze the correlation between each FRG and immune cells.ANO6, CS, EMC2, PANX1, and PIK3CA were found to be positively correlated with multiple immune cells such as CD8+ T cells, macrophages, neutrophils, and dendritic cells.CHAC1, SLC7A5, and SOCS1 were positively correlated with B cells, neutrophils, and dendritic cells.CHMP6 was nega- tively correlated with CD8+ T cells while positively correlated with CD4+ T cells.GPX4 was negatively correlated with B cells, CD8+ T cells, macrophage, neutrophil, and dendritic cells.G6PD was positively correlated with B cells (Supplementary Fig. S5-6 at https://ojs.ptbioch.edu.pl/index.php/abp/).
Due to the importance of immunotherapy, the differences in immune checkpoints between risk score subgroups were analyzed.The results indicated higher expression of most immune checkpoints including PDCD1, LAG3, etc. in the low-risk group, suggesting a greater benefit of immune checkpoint suppression therapy in low-risk patients.The high expression of CD80, PD-L2, and TNFSF4 in the high-risk group can aid in guiding the study for the optimization of immune checkpoint inhibitors (Fig. 11).

Neoadjuvant chemotherapy and drug sensitivity analysis
To investigate whether the risk score of tumors changes during neoadjuvant chemotherapy.We used two chemotherapy BC cohorts.A significant reduction in risk score was observed in patients treated with docetaxel and capecitabine by comparing paired patients from GSE18728 (Fig. 12A).In the GSE87455 cohort treated with epirubicin and docetaxel, 69 paired patients who received two cycles of treatment had significantly lower risk score than before treatment (Fig. 12B).Meanwhile, 57 paired patients who received six cycles of treatment had significantly lower risk score than before treatment (Fig. 12C).To further optimize chemotherapy in patients with BC, potential antitumor drugs were screened to compare the differences in drug sensitivity between the high-and low-risk groups.The results of sensitivity analysis showed that the IC50 values of drugs including AUY922, docetaxel, etoposide, imatinib, mitomycin C, paclitaxel, and SL 0101-1 were significantly reduced in the low-risk group, suggesting a higher sensitivity of the patients in the low-risk group to these drugs (Fig. 13A-G).Patients in the high-risk group were found to be more sensitive to AKT inhibitor VIII and cisplatin (Fig. 13H-I).

DISCUSSION
With the development of molecular diagnostic studies, medical professionals are increasingly resorting to the clinical practice of molecular diagnostics in BC, such as Oncotype dX and MammaPrint (Li et al., 2017;Barzaman et al., 2020).BC is a systemic disease and has To the best of our knowledge, this is the first study to explore the correlation between FRGs and cuproptosisrelated genes in patients with BC.Unlike other regulatory cell death processes, cuproptosis is primarily associated with adiposity of the tricarboxylic acid cycle and involves the loss of iron-sulfur cluster proteins and induction of ROS (Li et al., 2022;Tsvetkov et al., 2022).Gao and others (Gao et al., 2021) reported the accumulation of ROS due to copper retention, which inhibited the SLC7A11 levels, thereby enhancing oxidative stress and ferroptotic cell death in colorectal cancer.Thus, FRGs     may cause regulated cell death through the cuproptosisrelated mechanism.
In this study, the relationship between 259 FRGs and 13 cuproptosis-related genes in BC was comprehensively evaluated and screened for cuproptosis-related FRGs.The relationship between cuproptosis-related FRGs and OS in BC was analyzed and a novel prognostic model containing 11 cuproptosis-related FRGs was constructed.The accuracy of the model was verified by the ME-TABRIC cohort.Survival and ROC analyses revealed a good predictive ability of the model.Univariate and multivariate Cox analysis established the risk score of cuproptosis-related FRG signature to be an independent prognostic indicator for BC.The risk score was further found to be correlated with immune infiltrating cells and the expression of immune checkpoints between risk subgroups was analyzed.Finally, two neoadjuvant chemotherapy cohorts were used to show that the risk score was significantly reduced over the course of treatment, and nine antitumor drugs were screened for the treatment of patients with BC in different risk subgroups.
A significant difference between cuproptosis-related FRG signature and the clinical stage was observed, suggesting that higher risk scores were concentrated in more advanced tumor stages, ER-negative, PR-negative, and HER-2-positive, which further validates the poorer prognosis in the high-risk group.Meanwhile, the prognosis of cuproptosis-related FRG signature was confirmed in different subgroups with different clinical characteristics, except for male patients and patients with distant metastases, which could be because of the small sample size in these two subgroups.Therefore, the proposed model has demonstrated generalizability in different strata of patients with BC.
The cuproptosis-related FRG signature included ANO6, CHAC1, CHMP6, CS, EMC2, G6PD, GPX4, PANX1, PIK3CA, SLC7A5, and SOCS1.These genes included ferroptosis driver genes (ANO6, CHAC1, CS, EMC2, G6PD, PANX1, and PIK3CA) and ferroptosis suppressor genes (CHMP6, PGX4, and SOCS1).Currently, SLC7A5 is an unclassified regulator whose role in ferroptosis is unclear.There is growing evidence of the involvement of ANO6 in multiple forms of cell death, including necroptosis, pyroptosis, and ferroptosis (Ousingsawat et al., 2017;Ousingsawat et al., 2018;Simões et al., 2018).As a non-selective Ca 2+ -activated ion channel, lipid scramblase ANO6 plays a critical role in the induction of ferroptotic cell death by disrupting the stability of membrane phospholipids (Ousingsawat et al., 2019).Chen et al. reported that CHAC1 contributes to cystine starvation and thereby induces ferroptotic cell death in triple-negative BC cells via the GCN2-eIF2α-ATF4 pathway (Chen et al., 2017).Overexpression of CHAC1 has been linked to the enhanced sensitivity of prostate cancer cells to docetaxel, but the effect was reversed after co-treatment with ferroptosis inhibitors, suggesting that CHAC1 functions as a potential therapeutic target for castration-resistant prostate cancer by inducing ferroptotic cell death (He et al., 2021).Ferroptosis activators such as erastin and RSL3 increased CHMP6 accumulation by triggering calcium influx.Silencing of CHMP6 expression sensitized cancer cells to lipid peroxidationmediated ferroptosis but did not affect iron accumulation (Dai et al., 2020).CS increased the accumulation of citrate in hypoxic triple-negative BC cells, which promoted the migration and invasion of cancer cells (Peng et al., 2019).The overexpression of G6PD suggests a poor prognosis in hepatocellular carcinoma and inhibits ferroptotic cell death of the tumor by targeting cytochrome oxidoreductase (Cao et al., 2021).Luo et al. reported the resistance of G6PD for doxorubicin-induced apoptosis in triple-negative BC cells by the maintenance of high glutathione levels (Luo et al., 2022).Genetic models were used for the identification of GPX4 as a key regulator of ferroptosis and that inhibition of GPX4-induced ferroptosis enhanced the sensitivity of triple-negative BC cells to gefitinib (Seibt et al., 2019;Song et al., 2020).A recent study reported that high expression of PANX1 induced local immunosuppression in tumor microenvironment of basal-like BC, as evidenced by high infiltration levels of neutrophils and adenosine production (Chen et al., 2022).SLC7A5 supported the proliferation and growth of tumor cells by increasing mTORC1 activity through the upregulation of regulatory transcription factors under hypoxic conditions (Nachef et al., 2021).In addition, SLC7A5 inhibitors were used to treat patients with advanced biliary tract cancer such that no disease progression occurred for two years (Hayes et al., 2015).SOCS1 sensitizes cells to ferroptosis by the activation of p53 target gene expression and glutathione level reduction (Saint-Germain et al., 2017).
The results of GO analysis and KEGG pathway analysis revealed the relation of 125 cuproptosis-related FRGs to certain biological functions or pathways, such as response to oxidative stress, cellular response to chemical stress, autophagy, FoxO signaling pathway, and central carbon metabolism in cancer.Therefore, lipid peroxidation, a key to the induction of ferroptosis, is mediated by oxidative and antioxidant systems, along with autophagy (Chen et al., 2021).
Tumor-associated macrophages have been shown to promote tumor invasion in BC by promoting angiogenesis and remodeling the tumor extracellular matrix while evading the host immune response and suppressing the antitumor effects of cytotoxic T cells (Choi et al., 2018;Mehta et al., 2021).The results of this study revealed higher levels of macrophage infiltration in the high-risk group, accrediting the poor prognosis in the high-risk group to the immunosuppression of the tumor microenvironment.It was also found that patients in the low-risk group had a higher likelihood of benefitting from immunotherapy, while patients in the high-risk group may be more sensitive to immunosuppressive agents targeting CD80, PD-L2, and TNFSF4.Risk score was found to become lower after treatment through the neoadjuvant cohort, which may suggest a remission of the tumor.Finally, the sensitivity of high-and low-risk groups to different antitumor drugs was explored and results revealed that patients in the high-risk group were more sensitive to AKT inhibitor VIII and cisplatin.The results obtained from this study combined with the data available in the literature will aid the facilitation of the optimization of chemotherapy regimens for patients with BC.
However, the objects of this study were from public databases, and clinical data is essential for further validation.Although 125 cuproptosis-related FRGs have been identified, the correlation between FRGs and cuproptosis-related genes was not established.Finally, subgrouprelated pathways, immune infiltration, and drug sensitivity require further experimental validation.

Figure. 1
Figure.1 Flow chart of this study.

Figure 3 .
Figure 3. Construction of the prognostic signature based on cuproptosis-related FRGs in TCGA cohort.(A) Kaplan-Meier curves for OS of BC patients between high-risk and low-risk groups.(B) Time-dependent ROC curves for OS.(C) Distribution of survival status based on the risk score.(D) The high-risk and low-risk groups were divided based on the median risk score.(E) Heatmap showed the difference in expression of 11 cuproptosis-related FRGs in high-and low-risk patients.BC, breast cancer; FRGs, ferroptosis-related genes; OS, overall survival; ROC, receiver operating characteristic; TCGA, the cancer genome atlas.

Figure 4 .
Figure 4. Construction of the prognostic signature based on cuproptosis-related FRGs in METABRIC cohort.(A) Kaplan-Meier curves for OS of BC patients between high-risk and low-risk groups.(B) Time-dependent ROC curves for OS.BC, breast cancer; FRGs, ferroptosis-related genes; METABRIC, molecular taxonomy of breast cancer international consortium; OS, overall survival; ROC, receiver operating characteristic.

Figure 5 .
Figure 5. Cuproptosis-related FRG signature was shown to be an independent risk factor for OS in TCGA.(A) Univariate Cox regression analysis between OS and various prognostic parameters.(B) Multivariate Cox regression analysis between OS and various prognostic parameters.FRG, ferroptosisrelated gene; OS, overall survival; TCGA, the cancer genome atlas.

Figure 8 .
Figure 8. Kaplan-Meier curves of OS differences between high-and low-risk groups stratified by age, gender, T stage, N stage, M stage or TNM stage.OS, overall survival.

Figure 9 .
Figure 9. Construction of a prognostic nomogram.(A) Nomogram of OS prediction with age, T stage, N stage, M stage and signature as parameters.(B) The calibration curves of the nomogram for 3-, 5-, and 10-year OS prediction.OS, overall survival.