The analysis process of this study is shown in Figure 1 . The RNA sequencing data and corresponding clinical data of CM samples were downloaded from The Cancer Genome Atlas (TCGA) (http://cancergenome.nih.gov/) (23). The RNA sequencing data of normal skin samples were downloaded from the Genotype-Tissue Expression (GTEx) database (https://gtexportal.org/home/) (24). The expression values of all genes were publicly available and in level 3 HTseq fragments per kilobase of exon per million mapped fragments format. The merged RNA expression profile of the TCGA CM cohort (471 samples) and GTEx normal skin cohort (234 samples) was normalized, and batch effects between the TCGA and GTEX data were removed using the limma package in R (25). GSE65904 (26) and GSE72056 (27) were downloaded from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/) and used for molecular subtyping and single-cell validation. A total of 531 FAMGs were derived from 6 FAMG sets in the Molecular Signatures Database (http://www.gsea-msigdb.org/gsea/msigdb/) (28) and were listed in Supplementary Table 1 . And the clinical detail of TCGA samples were listed in Supplementary Table 2

To identify differentially expressed FAMGs (DEFAMGs), we analyzed the gene transcription data of the TCGA and GTEx cohorts using the limma package in R with a false discovery rate (FDR) < 0.05 and log2 fold change > 1. We used the pheatmap package in R to construct heatmaps of DEFAMGs. We used Gene Ontology (GO) (29) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway (30) analyses to identify pathways in the Database for Annotation Visualization and Integrated Discovery associated with DEFAMGs (31). We used Metascape (https://metascape.org) (32) to verify and visualize the functional enrichment of the DEFAMGs.

We adopted non−negative matrix factorization (NMF) consensus clustering (33) to divide all TCGA CM patients into 2 molecular subtypes using the R “NMF” package based on FAMGs expression pattern. Principal component analysis (34) was used to visualize FAM molecular subtypes by dimensionality reduction. Overall survival (OS) and progression-free survival (PFS) were compared between the FAM molecular subtype groups. Gene set enrichment analysis (GSEA) (35) was performed to assess differences in immune-related pathways between the molecular subtypes. In addition, we used ESTIMATE (36) and ssGSEA (37) algorithms to compare the immune cell infiltration and function between the molecular subtypes. To validate the reliability of the FAM subtypes, we also performed the abovementioned analyses using a GEO dataset (GSE65904) as an external verification dataset.

We identified the DEFAMGs between FAM molecular subtypes in the TCGA and GEO datasets and performed GO and KEGG analyses. The protein-protein interaction (PPI) network was built on the intersection of the TCGA and GEO cohorts using STRING (https://cn.string-db.org/) (38) and reconstructed in Cytoscape v3.6 (39) to select hub biomarkers.

The expression profiles of the FAMGs were analyzed using a weighted gene co-expression network analysis (WGCNA) network (40) to select gene modules that were highly associated with FAM molecular subtypes. Among the soft threshold values, the β with the highest mean connectivity (β = 3) was chosen. The module eigengene was associated with molecular subtypes. Modules with the highest correlation were selected, and the genes in these modules were identified as genes related to FAM molecular subtypes. The minimum module size was 25. The module with the threshold of membership>0.4 and P value <0.05 was identified.

To identify hub biomarkers of FAM molecular subtypes, we used a Venn plot to identify overlapping PPI genes, WGCNA module genes, and DEFAMGs with statistical significance in univariable Cox regressions in both the TCGA and GEO datasets. We used correlation-based network analysis to identify interactions among core biomarkers. We analyzed the prognostic value of the hub biomarkers in predicting OS, PFS, and disease-specific survival (DSS) in the TCGA and GEO cohorts.

To assess the immune-related and immune checkpoint–related pathways of the FAM biomarkers, we divided the patients in the TCGA CM cohort into groups with either high or low expression of each biomarker. The median expression level was used as the cut-off value. GSEA (41) was performed to identify significantly regulated pathways between the 2 groups. GSEA software was downloaded from the Broad Institute (http://www.broadinstitute.org/gsea/index.jsp). The gene sets of “c2.cp.kegg. v6.2. symbols”,”BIOCARTA_TCR_PATHWAY”,”REACTOME_PD_1_SIGNALING”,”LIN_TUMOR_ESCAPE_FROM_IMMUNE_ATTACK”,”WP_CANCER_IMMUNOTHERAPY_BY_PD1_BLOCKADE” and “WP_FERROPTOSIS” were downloaded from the Molecular Signatures Database (MSigDB). The normalized enrichment score (NES) was calculated for each gene set. Statistical significance was set at |NES| > 1, nominal P value < 0.05, and FDR q-value < 0.25.

We used the ESTIMATE algorithm to calculate the immune score, tumor purity, ESTIMATE score, and stromal score for each CM sample. For each core biomarker, the differences between the high- and low-expression groups were analyzed using violin plots. Furthermore, the association between the expression of FAM biomarkers and immune cell infiltration was assessed by Pearson correlation analysis. The Tumor Immune Estimation Resource (TIMER) database (https://cistrome.shinyapps.io/timer/) (42) was used to identify correlations between immune cells and FAM biomarkers.

The GSE72056 single-cell dataset (4,645 single-cell sequencing samples of Homo sapiens; platform: GPL18573 Illumina NextSeq 500) was used to verify the expression of the FAMGs in the TME and assess the relationship between immune cells and FAMGs. We used the Seurat package in R for batch calibration and data normalization (43). We used the t-SNE package (44) to perform cell cluster analysis and used the SingleR package (45) to identify cell subpopulation annotations. We verified the connections between immune cells and molecular subtype biomarkers by evaluating the biomarker expression level of each cell.

We assessed differences in the expression of 30 genes that previous studies suggested to be immune checkpoint–related genes (46–48) between groups with high or low expression of each FAM biomarker. In addition, we used Pearson correlation analysis based on the TIMER database to assess associations between the expression of FAM biomarkers and that of key immune checkpoints, including PD-1, programmed death-ligand 1 (PD-L1), and cytotoxic T-lymphocyte antigen 4 (CTLA-4). We used TIDE scores to assess differences in immunotherapy response between groups with high or low expression levels of each FAM biomarker.

We used the CellMiner database (https://discover.nci.nih.gov/cellminer) to construct an interaction network between chemotherapy sensitivity and FAM molecular subtype biomarkers. We performed Pearson correlation analysis to assess drug-gene associations.

The protein expression of the FAM biomarkers in normal and CM tissues was verified by analyzing immunohistochemical and immunofluorescent data extracted from The Human Protein Atlas (HPA; https://www.proteinatlas.org/).

Human melanoma A375 cell line was purchased from American Type Culture Collection (Manassas, VA); the M14, human immortalized keratinocyte (HaCaT), and normal human skin melanocyte (PIG1) cell lines with STR certification were purchased from Shanghai Guandao Biological Engineering Company (Shanghai, China); and the SK-MEL-28 cell line was purchased from the Chinese National Infrastructure of Cell Line Resource. All cells were cultured in Roswell Park Memorial Institute (RPMI) 1640 medium supplemented with 10% fetal bovine serum at 37°C in a 5% CO₂ atmosphere. Real-time quantitative polymerase chain reaction (qRT-PCR) was performed to determine relative gene expression levels. Total RNA was extracted with TRIzol reagent 24 hours. The concentration of RNA was determined by ultraviolet spectrophotometry. RNAs were reverse-transcribed into complementary DNAs (50 ng/µl) using a commercial complementary DNA reverse transcription kit. PCR with SYBR Green Master Mix (TaKaRa Bio, Kusatsu, Japan) was used to evaluate mRNA expression levels. Glyceraldehyde-3-phosphate dehydrogenase was used as the internal reference. All primers were synthesized by Servicebio Technology Co. (Wuhan, China). The primer sequences are shown in Supplementary Table 3 . The PCR program consisted of an initial denaturation at 95°C for 10 minutes followed by 45 cycles of 95°C for 10 seconds, 60°C for 30 seconds, and 72°C for 20 seconds. Target gene expression was calculated using the 2^−ΔΔCT method.

The survival, ggplot2, corrplot, pheatmap, singleR, and limma packages were executed using R software version 4.1.1 (https://www.r-project.org). A P value <0.05 was defined as statistically significant. The unpaired Student t-test was used to analyzing normally distributed continuous variables. Univariable Cox regressions and WGCNA were performed to identify FAMGs with significant prognostic values. Kaplan–Meier survival analyses and log-rank tests were used to assess OS, PFS, and DSS.

A total of 219 FAMGs were differentially expressed between CM tissues (n = 471) and normal skin tissues (n = 234), including 124 downregulated and 95 upregulated genes in the tumor samples ( Figures 2A, B ). GO analysis showed that the DEFAMGs were mainly enriched in lipid and FAM-related biological processes, cellular components, and molecular functions ( Figure 2C ). KEGG analysis showed that the DEFAMGs were mainly enriched in metabolic pathways and FAM pathways (arachidonic acid, linolenic acid, glycerophospholipid, etc.) ( Figure 2D ). Metascape functional annotation confirmed that DEFAMGs were mainly enriched in FAM processes ( Figure 2E ).

Univariable Cox regression revealed that 49 of the 219 DEFAMGs were significantly correlated with OS ( Figure 3A ). The NMF algorithm divided the TCGA CM samples into 2 clusters (cluster 1 and 2) ( Figure 3B ). The relationship between cophenetic, dispersion, and silhouette coefficients showed that the 2 clusters were significantly different ( Figure 3C ), which was further verified by principal component analysis ( Figure 3D ). Survival analysis showed that cluster 2 patients had better OS and DFS than cluster 1 patients did ( Figures 3E, F ). The GSEA results indicated that the immune response and immune system process were upregulated in cluster 2 ( Figures 3G, H ). Given the higher immune level and better prognosis in cluster 2, we compared the TME between the two clusters and found that immunocyte subpopulations were highly enriched in cluster 2 ( Figure 3I ). Compared with cluster 1, cluster 2 had significantly higher stromal, immune, and ESTIMATE scores ( Figure 3J ). The ssGSEA showed that cluster 2 also had higher levels of immune infiltration and more activated immune functions ( Figure 3K ). These findings were consistent with those of identical analyses performed using a GEO dataset (GSE65904) as an external validation dataset ( Supplementary Figures 1A–J ).

We identified 55 DEFAMGs between the 2 clusters of the TCGA CM samples ( Figure 4A ), including 22 upregulated and 33 downregulated genes in cluster 2 ( Figure 4B ). GO and KEGG analyses revealed that the DEFAMGs were enriched in the oxidation-reduction process, arachidonic acid metabolism, and metabolic pathways ( Figure 4C ). Using GEO samples, we identified 76 DEFAMGs between the 2 clusters ( Figure 4D ), including 48 upregulated and 28 downregulated genes in cluster 2 ( Figure 4E ). GO and KEGG analyses showed that these DEFAMGs were enriched in the oxidation-reduction process, the lipid metabolic process, and metabolic pathways ( Figure 4F ). The PPI network included 28 downregulated and 36 upregulated DEFAMGs ( Figure 4G ).

We extracted the expressions of 531 FAMGs in TCGA samples and GSE65904 samples for WGCNA. Seven modules were identified by the average linkage hierarchical clustering based on the soft-thresholding power in TCGA samples, and the WGCNA traits heat map showed that the Module Eigengenes (ME) turquoise was selected ( Figure 5A ). The Module Eigengenes (ME) turquoise had the highest correlation with the FAM clusters (|cor| =0.59, P-value =5e-44) and contained 110 FAMGs. The dendrogram of genes clustered according to a dissimilarity measure (1-Topological Overlap Matrix, TOM) was shown in Figure 5B . The Sample dendrogram and soft threshold of WGCNA according to FAM Molecular subtypes were shown in Figures 5C, D . The co-expression network was constructed, and 3 modules were determined in GEO samples. Correlation analysis between the module eigengenes and FAM cluster showed that the ME turquoise ( Figures 5E–H , Module–trait relationships = 0.39, P = 0.000) had the highest association with the FAM clusters. 92 genes in the module were considered to be hub FAM-related Modules genes. The overlapping part of the Venn plots identified 6 genes (i.e. ACSL5, ALOX5AP, CD1D, CD74, IL4I1 and TBXAS1) as core FAM biomarkers of CM ( Figure 5I ).

Kaplan-Meier analyses showed that higher expression of ACSL5, ALOX5AP, CD1D, CD74, IL4I1, or TBXAS1 was associated with better OS, PFS, and DSS in the TCGA cohort, and with better OS in GEO cohorts ( Figures 6A–D ). In addition, univariable and multivariable Cox regressions showed that the 6 genes were independent prognostic factors for OS ( Supplementary Table 4 ).

GSEA revealed that antigen processing and presentation, T-cell receptor, the PD-1 signaling pathway, cancer immunotherapy by PD-1 blockade, tumor escape, and ferroptosis were all positively correlated with the expression levels of the 6 FAMGs ( Figures 7A–F ). These results indicate that the 6 genes may be involved in tumor immune regulation, immune evasion, and tumor cell response to PD-1 blockade.

Compared with the groups with low expression of the 6 FAMGs, those with high expression of the 6 FAMGs had significantly higher ESTIMATE, immune, and stromal scores but lower tumor purity ( Supplementary Figure 2 ). The expression levels of the 6 FAMGs were significantly positively correlated with the ESTIMATE, immune, and stromal scores but negatively correlated with tumor purity ( Supplementary Figure 3 ). The expression levels of the 6 FAMGs were also positively correlated with the infiltration of B cells, CD8⁺ T cells, CD4⁺ T cells, macrophages, neutrophils, and dendritic cells ( Figure 8 ).

The single-cell analysis confirmed that the expression levels of the 6 FAMGs were relatively high in the CD8⁺ T cell, B cell, and natural killer (NK) cell subsets ( Figures 9A, B ). The average and percent expressions of the 6 FAMGs are illustrated in a bubble chart of single-cell data ( Figure 9C ). The expression levels of the 6 FAMGs are illustrated in a violin chart of single-cell data ( Figure 9D ).

A total of 30 immune checkpoints, including PD-L1, PD-1, and CTLA-4, were found to be differentially expressed between groups with low or high expression of the 6 FAMGs ( Supplementary Figure 4 ). The expression levels of the 6 FAMGs were significantly correlated with those of PD-L1, PD-1, and CTLA-4 ( Figures 10A–C ). A comparison of the groups’ TIDE scores demonstrated that the group with high expression of the 6 FAMGs had better immunotherapy responses ( Figure 10D ).

Chemotherapy sensitivity analysis showed that positive expression of the 6 FAMGs was associated with high sensitivity to certain chemotherapy drugs ( Figure 11A ). Pearson correlation analysis confirmed that higher expression levels of the 6 FAMGs were associated with higher sensitivity to some chemotherapy and targeted drugs, such as trametinib, selumetinib, cobimetinib, carboplatin, oxaliplatin, cisplatin, and dacarbazine ( Figure 11B ).

Semiquantitative analyses of immunohistochemical images confirmed the expression levels of the 6 FAMGs were significantly higher in CM tissues than in normal skin ( Figure 12A ). Immunofluorescence analysis revealed that ACSL5, CD1D, CD74, IL4I1, and TBXAS1 were mainly located in the cytosol and nucleoplasm ( Figure 12B ). qRT-PCR analysis revealed that the 6 FAMGs had significantly upregulated expression in A375, m14, and SK-MEL-28 cell lines compared with HaCaT and PIG1 cell lines ( Figure 12 ).

This study comprehensively analyzed the role of FAMGs in the subtypes, prognosis, immune infiltration and immunotherapy response in CM. We first systematically investigated the correlation between DEFAMGs and the prognosis of patients with CM. Then we applied the NMF method to deconvolute the FAMGs expression profiles and identified 2 FAM clusters, cluster 1 and cluster 2, based on the TCGA CM cohort and further validated using the GEO cohort. According to the FAM subtypes, we identified 6 hub genes (ACSL5, ALOX5AP, CD1D, CD74, IL4I1, TBXAS1) as FAM subtype biomarkers. Finally, we verified the expression levels of the 6 genes in the HPA database and the qRT-PCR. Overall, our findings show that ACSL5, ALOX5AP, CD1D, CD74, IL4I1, and TBXAS1 are key FAM biomarkers that affect the function and phenotype of infiltrating immune cells and are associated with immunosuppression in the TME.

Tumor molecular subtypes that are based on predicted prognosis and TME composition have become a hotspot in cancer research, but few studies have provided comprehensive insight into the integral role of FAM in tumor molecular subtypes, especially in CM. For example, Ying et al. used a method of consensus clustering to build specific FAM-related molecular subtypes that were associated with malignancy and prognosis in glioma (49), but they did not explore the differences in immunocyte infiltration into the TME between the FAM-related molecular subtypes. Samson et al. constructed a mutational subtype of melanoma based on the BRAF mutation patterns of 2 FAM-related genes, ALDH1A1 and ALDH1A3; however, this CM subtype accounts for ALDH1A3 expression as only a prognostic marker for BRAF/MEK inhibitor treatment response in BRAF-mutant metastatic melanoma patients (50). In the present study, we identified 2 FAM subtypes (cluster 1 and cluster 2) with different immune infiltration patterns based on the FAMG phenotype of CM. GSEA showed that immune response and immune system process were significantly regulated in FAM cluster 2. Immunocyte subsets, such as CD8⁺ T cells, B cells, tumor-infiltrating lymphocytes, T follicular helper cells, regulatory T cells, and T helper cells, were highly enriched in FAM cluster 2, indicating that cluster 2 had higher levels of T-cell subset infiltration than cluster 1 did.

T cells degrade fatty acids through fatty acid oxidation to acquire lipids for energy. In addition, the development and function of different T-cell subsets are closely linked to the balance between fatty acid synthesis and fatty acid oxidation in the TME. For example, the preferential usage of fatty acid oxidation has been linked to the development of CD8⁺ memory T cells and the induction of CD4⁺ regulatory T cells over other T-cell lineages. Especially in CD4⁺ T-cell subsets, the FAM profile changes during an immune response and is influenced by different tissues and different types of inflammation. Our FAM subtypes provide further evidence that different FAM patterns affect the abundance of T-cell subsets in the CM TME. Recent studies confirmed that T-cell subpopulations are the targets of ICI and chimeric antigen receptor T-cell therapy and that the status of T cells can strongly influence patients’ prognosis (51, 52). Our results also showed that FAM cluster 2, with its high levels of T-cell infiltration, was associated with favorable outcomes, which suggests that cluster 2 may be more responsive to immunotherapy than cluster 1. In addition, our GO and KEGG analyses showed that the DEFAMGs between cluster 2 and cluster 1 were mainly enriched in the oxidation-reduction process, peroxisome proliferator-activated receptor (PPAR) signaling pathway, and lipid metabolic pathways. Accumulating evidence shows that the PPAR signaling pathway, which is involved in lipid metabolism, energy homeostasis maintenance, inflammation, and immune tolerance, has a role in carcinogenesis. Lee et al. found that AMP-activated kinase induces the upregulation of CPT1 and PGC-1 to activate the PPAR signaling pathway, which may more indirectly favor fatty acid oxidation (53). Therefore, the PPAR signaling pathway may be a key determinant of FAM patterns between FAM subtypes, and it may be used in the context of novel therapeutic strategies against CM.

Based on the FAM molecular subtypes, we identified the 6 FAMGs (ACSL5, ALOX5AP, CD1D, CD74, IL4I1, TBXAS1) that play vital regulatory roles in the melanoma TME. Among the 6 hub FAMGs, IL4I1 (interleukin-4-induced gene 1) has a vital role in immunosuppressive functions and tumor immune escape. In 2009, Carbonnelle et al. were the first to propose that the novel immunosuppressive enzyme IL4I1, which is produced by the neoplastic cells of several B-cell lymphomas and by tumor-associated macrophages, is a prognostic biomarker and therapeutic target in cancer (54). Later, Fanette et al. found that tumor-associated macrophages with high IL4I1 expression can inhibit T-cell proliferation in vitro through H₂O₂ production. They also confirmed that in human melanoma and mesothelioma, minimal IL4I1 activity-induced tumor escape was preceded by a rapid diminution of interferon γ–producing cytotoxic antitumor CD8⁺ T cells (55). Cousin et al. further showed that IL4I1 stimulates the generation of Foxp3⁺ regulatory T cells and limits T helper 1 and T helper 2 polarization in vitro, and their findings reinforced the concept that IL4I1 facilitates tumor escape from the immune response (56). Sadik et al. recently identified IL4I1 as a major aryl hydrocarbon receptor (AHR)-activating enzyme that promotes AHR-driven cancer cell motility and suppresses adaptive immunity. Compared with IDO1 or TDO2, IL4I1 had a stronger effect on AHR activity, which suggests that IL4I1 is an alternative metabolic immune checkpoint that could be therapeutically targeted in patients in whom combined ICI and IDO1 inhibition therapy has failed (57).. In the present study, we consistently observed that the PD-1, T-cell receptor, and tumor escape from immune attack signaling pathways were significantly enriched in CM samples with high IL4I1 expression. The differential expression of IL4I1 was significantly correlated with the infiltration of immune cells such as B cells, CD8⁺ T cells, macrophages, and neutrophils in the TME. These findings suggest that IL4I1 can be an indicator of response to anti–PD-1 treatments and is a novel metabolic immune checkpoint in CM.

Previous studies showed that ACSL5 (acyl-CoA synthetase 5), a nuclear-coded FAMG expressed in the mitochondria, can convert carbons from citrate to bioactive fatty acids to stimulate the inflammatory response in the TME (58, 59). For example, Klaus et al. found that the high activity of ACSL5 enhances caspase-3 and caspase-7 activity to promote apoptosis regulated by TP53 status via WNT2B palmitoylation in enterocytes and colorectal adenocarcinomas (60). In addition, ACSL5 can be used to predict survival and immunotherapy response. Gassler et al. showed that lower ACSL5 expression is a prognostic marker for early recurrence in patients with colorectal adenocarcinoma (61). Chen et al. showed that breast, colorectal, lung, or ovarian cancer patients who have higher ACSL5 expression have good survival outcomes (62). Our results showed that ACSL5 is an important indicator of the OS, PFS, and DSS of CM patients. We also found that ACSL5 is correlated with the ferroptosis signaling pathway in CM, which suggests that ACSL5 can inhibit the proliferation of tumor cells by inducing ferroptosis.

ALOX5AP is a key enzyme that facilitates the activity of 5-lipoxygenase (5-LOX), which metabolizes arachidonic acid to leukotrienes. Moore et al. found that the 5-LOX/ALOX5AP pathway can affect cancer-related immune evasion in the TME (63). Ye et al. confirmed that ALOX5AP is involved in M2 macrophage recruitment, infiltration, and polarization, which can indirectly block tumor-specific T-cell activity to promote immune evasion in the ovarian cancer TME (64).

CD1D, a major histocompatibility complex class I–like molecule, presents lipoidal antigen to NK T cells, which are involved in the innate anticancer immune response. Bernal et al. confirmed that the CD1D molecule plays a crucial role in the induction of melanoma immune evasion by forming a complex with β2 microglobulin in the CD1D/NK T-cell axis (65). CD1D⁺ tumors can evade recognition by NK T cells in the CD1D/NK T-cell axis by shedding glycolipids, which presumably replace the endogenous lipids that are bound to CD1D and cannot be recognized by NK T cells (66).

CD74 has been implicated to play a tumor-progression role in the immune microenvironment of CM patients. Figueiredo et al. confirmed that CD74, as a macrophage migration inhibitory factor (MIF), regulates the activity of macrophages and other immune cells via the CD74-MIF signaling pathway. Interfering with MIF-CD74 immunosuppressive signaling can restore the antitumor immune response in metastatic melanoma (67). According to other studies, the polarization of macrophages to M2 macrophages induces immunosuppression by suppressing cytotoxic T cells in the TME (68). In the present study, we also showed that CD74 is positively associated with the infiltration of macrophages, dendritic cells, and neutrophils in CM. Our single-cell analysis of the TME further showed that high CD74 expression was enriched in CD8⁺ T cells, CD4⁺ T cells, macrophages, neutrophils, and dendritic cells in the CM TME.

In the arachidonic acid cascade, TBXAS1 encodes for thromboxane synthase, which converts prostaglandin H2 into thromboxane A2, a process that involves the modulation of cell cytotoxicity and tumor growth and metastasis (69, 70). Abraham et al. showed that TBXAS1 genes are associated with breast cancer risk (71). We found that TBXAS1 is significantly correlated with neutrophils in the TME, indicating that high TBXAS1 expression upregulates the inflammation response in the TME.

Although it presents encouraging results, our study had several limitations. First, the FAM tumor subtype was constructed and validated using open data sources (TCGA and GEO). Testing the subtype in an independent patient cohort would validate the reliability of our subtypes in identifying CM patients. Second, our study lacked in vivo experiments to verify the molecular function of the 6 hub genes. Future studies are needed to investigate the mechanisms underlying the 6 FAM biomarkers’ mediation of CM progression and the immune microenvironment.