FTO–CHRM3 axis regulates multiple myeloma progression: a machine learning-based identification

Abstract

Background

Fat Mass and Obesity-Associated (FTO) is linked to multiple myeloma (MM) progression, but its action mechanisms are poorly understood.

Methods

Machine learning and correlation analysis identified genes with consistent expression patterns as FTO-associated genes. Additionally, the expression levels of FTO and its associated genes in MM tissues were validated through immunohistochemistry. The impact of FTO-associated genes expression on the immune microenvironment of MM was assessed through immune infiltration analysis. Single-cell RNA sequencing (scRNA-seq) elucidated cellular expression patterns.

Results

FTO was found to be significantly upregulated in MM and associated with poor prognosis. Four potential FTO-associated genes (CHRM3, GINS3, RRM2, and SHCBP1) were identified. Correlation and immunohistochemical analyses further focused the FTO-CHRM3 axis. Functional enrichment analysis unveiled that FTO and CHRM3 were involved in “DNA replication” and “cell cycle”. The infiltration of M2 macrophages and eosinophils were altered in high FTO/CHRM3 expression groups. ScRNA-seq highlighted the monocyte-specific expression of SLC8A1 in the calcium signaling pathway in MM.

Conclusion

The FTO-CHRM3 axis contributes to MM progression via calcium signaling pathway dysregulation, and the exploration of its underlying mechanisms provides insights for targeted MM interventions.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00277-026-06940-2.

Introduction

Multiple myeloma (MM), the second most prevalent hematologic malignancy, is characterized by the clonal proliferation of malignant plasma cells within the bone marrow. Over the past decades, the therapeutic landscape for MM has been revolutionized through the introduction of proteasome inhibitors, immunomodulatory agents, monoclonal antibodies, CAR-T therapy, and autologous stem cell transplantation. These advances have led to marked improvements in patient survival and clinical outcomes [1]. However, the emergence of drug resistance and disease relapse continues to pose major clinical challenges, and MM remains an incurable condition. It is increasingly recognized that the progression of MM is driven by a multifactorial process, principally involving genetic abnormalities, remodeling of the bone marrow microenvironment, and epigenetic alterations [2]. Nevertheless, the etiology of MM remains poorly understood, underscoring the imperative for ongoing investigation into its molecular mechanisms and the discovery of novel therapeutic targets.

N6-methyladenosine (m6A) is the most abundant and conserved internal RNA modification in eukaryotic organisms, particularly within higher eukaryotes. Over the past few decades, m6A-mediated biological functions have been increasingly implicated in the progression of various tumor types [3]. It is dynamically regulated by methyltransferases (“writers”) and demethylases (“erasers”) such as Fat Mass and Obesity-Associated (FTO) and ALKBH5 [4]. Initially linked to human obesity through genome-wide association studies in 2007, the FTO gene was identified as a major genetic risk factor due to intronic variants [5]. More recently, studies have demonstrated that FTO is highly expressed in MM patients, where its overexpression correlates with poor clinical outcomes. Mechanistically, FTO acts as an oncogene in MM by enhancing proliferation, migration, and invasion of malignant cells via YTHDF2-dependent regulation of the HSF1/HSP signaling pathway [6]. Furthermore, FTO is highly expressed in drug-resistant MM cells, where it downregulates SOD2 in an m6A-dependent manner. This leads to significant accumulation of intracellular reactive oxygen species (ROS), thereby activating the canonical NF-κB signaling pathway and promoting drug resistance [7]. The m6A demethylase FTO is closely associated with the advancement of MM, yet its action mechanisms remain incompletely elucidated.

In this study, a series of advanced bioinformatics approaches combined with public database resources were employed to screen for FTO-associated genes. This study comprehensively elucidated the mechanistic roles of FTO and its related pathway in the progression of the disease, providing new perspectives for understanding the pathogenesis of MM.

Materials and methods

Data collection and study design

GSE118985 (GPL570), which was served as training cohort 1 and encompassed 750 bone marrow tissue transcriptome data samples [682 MM and 68 control samples]. Additionally, the MMRF-COMMPASS dataset (Illumina) was sourced from the UCSC Xena platform (https://xenabrowser.net/), and served as training cohort 2. It featured transcriptome data from 859 MM bone marrow samples along with their corresponding clinical features and follow-up information. The datasets GSE118985, GSE24870, GSE24080, and GSE124310 were retrieved from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo). Specifically, the datasets GSE24870 (GPL571) and GSE24080 (GPL570) functioned as validation cohorts 1 and 2, respectively, with 43 (MM: control = 23:20) and 559 samples (all MM samples) of bone marrow tissue transcriptome data (Table S1). The scRNA-seq dataset GSE124310 (GPL16791) comprised 32 human bone marrow tissue samples, divided into 9 control samples and 23 MM samples. Figure S1 was the flow diagram of this research methodology.

Statistical analysis

All analyses were conducted in R version 4.4.1, with inter-group differences assessed using Wilcoxon test or log-rank test, establishing a significance level at P < 0.05. For specific methods, please refer to the supplementary materials 1.

Result

The impact of FTO gene expression levels and its molecular mechanisms on MM

Initially, the expression patterns of the FTO gene were examined in training cohort 1 and validation cohort 1. The results indicated significant expression differences between the MM and control groups, with an upregulation trend observed in the MM group (Fig. 1A). This suggested that FTO might be crucial in the development and progression of MM. KM curves revealed that patients with high FTO expression had a significantly lower survival probability contrasted to those with low expression (Fig. 1B). The expression of FTO gene underlying molecular mechanisms by analyzing the correlation between FTO expression levels and patient outcomes, as well as the associated biological pathways. This finding underscores the critical role of FTO expression in shaping patient prognosis. Further investigation uncovered a significant differences between FTO expression levels and both patient age (> 65 or ≤ 65) and disease stage (stage Ⅰ~Ⅲ) (Fig. 1C). Additionally, chi-square tests highlighted significant disparities in the distribution of age and stage subtypes between the high and low FTO expression cohorts (Fig. 1D). Drug sensitivity analysis showed significant differences in the sensitivity to 189 drugs between the high and low FTO expression groups (adj.P < 0.05, Table S2). Notably, drugs such as camptothecin, vinblastine, and cisplatin exhibited differential sensitivity, suggesting that FTO expression levels may influence the therapeutic response to these agents.

Fig. 1.

Related studies on Fat Mass and Obesity-Associated (FTO). A boxplot showing the expression differences of FTO between multiple myeloma (MM) and controls. The left panel represents training set 1, and the right panel represents validation set 1. B Kaplan-Meier (KM) curves for high and low FTO expression groups. C analysis of differences in FTO gene expression levels among clinical characteristics. D distribution of clinical characteristic subgroups between high and low FTO expression groups. E gene set enrichment analysis (GSEA) of FTO. F gene set variation analysis (GSVA) between high and low FTO expression groups. ns, no significance; * P < 0.05; ** P < 0.01; **** P < 0.0001

To elucidate the molecular mechanisms underlying FTO’s involvement in MM, we conducted GSEA. The analysis identified 15 pathways significantly enriched in association with FTO expression, including “cell cycle”, “ubiquitin-mediated proteolysis”, “mismatch repair”, and “spliceosome” (Fig. 1E). Subsequent GSVA analysis provided deeper insights into the activation and suppression of multiple biological pathways in the high FTO expression group. Notably, pathways such as “regulation of autophagy”, “aminoacyl-tRNA biosynthesis”, “nucleotide excision repair”, and “inositol phosphate metabolism” were found to be activated, whereas “cytokine-cytokine receptor interaction”, “complement and coagulation cascades”, “MAPK signaling pathway”, and “chemokine signali (Fig. 1F, Table S3).

Identification and functional exploration of methylation-related genes

Differential analysis between FTO two groups identified 320 DEGs in the training cohort 2, comprising 112 upregulated and 208 downregulated genes (Figures S2A). Further correlation analysis pinpointed 36 FTO-associated genes (Figure S2B). To elucidate the molecular functions and mechanisms of these methylation-related genes, enrichment analysis was conducted. GO enrichment analysis yielded 224 entries, with 170 significantly enriched in biological processes (BP), 14 in cellular components (CC), and 40 in molecular functions (MF). The primary enriched items were “positive regulation of DNA metabolic process”, “G protein-coupled peptide receptor activity”, “C-C chemokine receptor activity”, and “establishment of protein localization to peroxisome” (Figure S2C). KEGG enrichment analysis highlighted 10 significantly enriched pathways, including “insulin secretion”, “cell adhesion molecules”, and “viral myocarditis” (Figure S2D). These pathways were involved in cell signaling, metabolic regulation, and immune response, potentially linking methylation to MM. To visually represent the interactions among these FTO-associated genes, PPI network was constructed, containing 36 nodes and 16 protein interaction pairs. RRM2 and MKI67 were observed to have multiple interaction pairs with other proteins (confidence score ≥ 0.4) (Figure S2E).

Identification of four potential key FTO-associated genes

To identify the key FTO-associated genes MM, univariate Cox regression analysis and PH assumption tests were executed on the 36 methylation-related genes, resulting in the identification of 27 survival-related genes (Fig. 2A, Table S4). Further analysis revealed that four genes exhibited significant expression differences and consistent trends in training cohort 1 and validation cohort 1. Specifically, CHRM3 was upregulated in MM, while GINS3, RRM2, and SHCBP1 were downregulated (Fig. 2B). KM curves demonstrated CHRM3, GINS3, RRM2, and SHCBP1 lower survival probabilities in the high expression groups (Fig. 2C-F). To validate the reliability of these genes, 101 different combinations of machine learning algorithms were employed to construct predictive models, with RSF+CoxBoost emerging as the optimal model (Fig. 2G). All genes passed the screening and were confirmed as key FTO-associated genes, including CHRM3, GINS3, RRM2, and SHCBP1.

Fig. 2.

Identification of potential FTO-associated genes. A univariate Cox regression analysis of 36 FTO-associated genes. B in training set 1 and validation set 1, four genes among the 27 survival-related genes showed significant differences between MM and controls with consistent trends. C-F. KM curves for the four genes. G. machine learning combination screening for the optimal model, with these four genes selected as potential key FTO-associated genes. * P < 0.05; ** P < 0.01; *** P < 0. 001; **** P < 0.0001

Determine the FTO-CHRM3 axis in MM

Given the positive correlation between CHRM3 and FTO (cor = 0.355), as well as CHRM3 was upregulation in MM, the potential mechanisms by which the FTO-CHRM3 axis affects MM were investigated. To this end, immunohistochemistry was employed to scrutinize the protein expression levels of CHRM3 and FTO in bone marrow biopsies from patients with MM. Our findings revealed that both CHRM3 and FTO were significantly upregulated in MM tissues contrasted to normal controls (Fig. 3).

Fig. 3.

Immunohistochemical results of CHRM3 and FTO between MM and controls. A immunohistochemical results of CHRM3 between MM and controls. B immunohistochemical results of FTO between MM and controls. ***, P < 0. 001

Clinical correlation and functional analysis of FTO-CHRM3 axis

The link between the expression level of CHRM3 and clinical features was further investigated. The results ascertained that the expression level of CHRM3 significantly differed between age and disease stage groups (Fig. 4A). Chi-square test results also ascertained significant differences in the proportions of CHRM3’ expression groups in age and stage subgroups (Fig. 4B). Drug sensitivity tests revealed that the sensitivity to multiple drugs, including camptothecin, vinblastine, and cisplatin, showed significant differences among the expression groups of CHRM3 (Fig. 4C, Table ).

Fig. 4.

Functional studies on CHRM3. A analysis of differences in CHRM3 expression levels among clinical characteristics. B distribution of clinical characteristic subgroups between high and low CHRM3 expression groups. C drug sensitivity analysis of CHRM3. D GSEA of CHRM3. E GSVA between high and low CHRM3 expression groups. ns, no significance; * P < 0.05; ** P < 0.01

Notably, GSEA showed that CHRM3 was enriched in “DNA replication”, “spliceosome”, “pyrimidine metabolism”, and “cell cycle” (Fig. 4D). Further exploration of the pathways between different expression groups of CHRM3 was conducted using GSVA. Multiple pathways were activated in the CHRM3 high expression groups, including “nucleotide excision repair”, “DNA replication”, “proteasome” and “RNA polymerase”. Multiple pathways were inhibited, including “complement and coagulation cascades”, “intestinal immune network for iga production”, “intestinal immune network for IGA production” and “ribosome” (Fig. 4E). These findings provided important evidence for understanding the biological functions of CHRM3 in MM.

Functional interaction and epigenetic regulation of FTO and CHRM3

To better comprehend the function and connections of the FTO-CHRM3 axis, GeneMANIA network analysis was performed. The network included a total of 20 interacting genes and 167 interaction pairs. Specifically, CHRM3 colocalized with ENTPD3 and had physical interactions with JAKMIP1 and MAGED2. Additionally, FTO had a physical interaction with TMEM147 and was coexpressed with HTR7. Enrichment was observed in “postsynaptic signal transduction”, “G protein-coupled amine receptor activity”, and “acetylcholine receptor signaling pathway” (Figure S3A). To further understand the specific subcellular localization of FTO and CHRM3 within cells, subcellular protein localization analysis was conducted using HPA. FTO was primarily localized in the cytosol and possibly in vesicles, while CHRM3 was localized in the plasma membrane FigureS3B). Under high-confidence conditions, CHRM3 had four methylation sites located at positions 1596, 2507, 2797, and 4936 (Figure S3C).

FTO regulation of CHRM3 affects the calcium signaling pathway

In training cohort 2, KM curves showed the highest significance level between the FTO+CHRM3 + and FTO-CHRM3- groups (P < 0.001), thus these two groups were selected for differential analysis (Fig. 5A). 518 upregulated and 465 downregulated genes were among the 983 DEGs that were found (Fig. 5B). To learn more about the biological roles of these DEGs, enrichment analysis was carried out. GO enrichment analysis yielded 1588 entries, with 1326 significantly enriched in BP, 118 in CC, and 144 in MF (Fig. 5C). KEGG pathway enrichment analysis unveiled 62 significantly enriched pathways, including “cytokine-cytokine receptor interaction”, “Th1 and Th2 cell differentiation”, and “IL-17 signaling pathway” (Fig. 5D). GSEA further indicated that CHRM3 was contained in the calcium signaling pathway (Fig. 5E, Table S5). Therefore, the calcium signaling pathway was selected as the key pathway for subsequent in-depth analysis.

Fig. 5.

Conduct enrichment analysis on the differentially expressed genes (DEGs) screened out from the optimal combination. A KM curve of different expression groups of FTO-CHRM3. B volcano map of DEGs between the FTO+CHRM3 + and FTO-CHRM3- groups. C Gene Ontology (GO) enrichment analysis of DEGs. BP, Biological Process; MF, Molecular Function; CC, Cellular Component. D Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis of DEGs. E calcium signaling pathway in GSEA

ScRNA-seq analysis reveals the gene expression patterns of FTO-CHRM3-calcium signaling pathway axis in MM

To better understanding of the mechanisms of action of FTO-CHRM3 axis in MM at the single-cell level, the GSE124310 dataset was analyzed. After data filtering, a total of 22,837 cells and 20,849 genes were obtained (Figure S4A). Subsequently, the top 2,000 genes with the highest variation were screened out, and the top 30 principal components were extracted based on these highly variable genes (Figure S4B-C). Through clustering analysis, these cells were ultimately divided into 12 different cell clusters (Figure S4D). By comparing with known marker genes, seven types of cells were successfully annotated, namely T cells, monocytes, NK cells, erythrocytes, B cells, plasma cell, and dendritic cells (Figure S5A). After determining the cell types, the focus was shifted to the FTO gene, the CHRM3 gene, and the genes related to the key pathway (calcium signaling pathway) (Table S6), which were considered key genes in order to reveal their specific roles in cell function through subsequent analysis. Among them, six genes (CACNA1B, CACNA1E, ERBB4, ATP2B2, GRM1, and MYLK2) were not present in the dataset, and ultimately 17 key genes were retained for further in-depth study.

To comprehensively understand the expression patterns of these key genes in different cell types, their expression and distribution in various cells were visualized. The results unveiled that these key genes were expressed in different cell types (Figure S5B), among which ATP2B4 and PRKX were expressed in NK cells. When further analyzing the expression differences of key genes between different groups, it was found that only the SLC8A1 gene showed significant expression differences in the monocyte group (Figure S5C). Based on this finding, monocytes were identified as the key cell type for pseudotime analysis to explore the dynamic changes of cells over time. The analysis results showed that the cell differentiation process could be divided into three states. The differentiation trajectories of the control and MM groups were compared, with the primary differences observed at stage 2 (Figure S5D). By analyzing the expression of key genes on the time trajectory, it was found that the expression levels of these key genes were relatively higher in the early stage of monocyte differentiation (state 1) in MM samples (Figure S6A). As cell differentiation progressed, these key genes were expressed in the middle and late stages (state 2 and state 3), and their expression was more significant in the late stage of differentiation (state 3) (Figure S6B). Additionally, the interactions between cells were also analyzed (Figure S6C). Compared with the control group, it was found that the interaction strength between monocytes and dendritic cells in the MM group increased, while the interaction strength between NK cells and T cells decreased, provide a fresh viewpoint for comprehending their MM mechanisms of action.

Discussion

MM is a common hematologic malignancy driven by genetic abnormalities and bone marrow microenvironment remodeling. FTO is an Fe²⁺- and α-ketoglutarate (2-OG)-dependent dioxygenase belonging to the RNA demethylase family. Its primary function is to catalyze the demethylation of m6A in RNA molecules, converting m6A into hm6A via oxidation and ultimately reducing it to adenosine, thereby dynamically regulating m6A modification levels [8, 9]. In MM, FTO overexpression enhances the expression stability of oncogenes such as MYC by reducing their m6A modification, promoting tumor proliferation and drug resistance [8]. Inhibitors targeting FTO can restore m6A modification levels, suppress MM cell proliferation, and induce apoptosis [10]. Moreover, interactions between MM cells and the bone marrow microenvironment (e.g., BMSCs, adipocytes) suggest that FTO may influence MM cell nutrient utilization efficiency by regulating the expression of genes involved in glutamine and fatty acid metabolism, thereby enhancing drug resistance and survival [11]. Thus, we propose that FTO, as a key regulator of m6A modification, promotes malignant progression in MM through a core mechanism of “demethylation-mediated stabilization of oncogenes”.

Following the identification of the FTO-CHRM3 axis via correlation and immunohistochemical analyses, functional enrichment revealed their joint involvement in the cell cycle pathway. The cell cycle pathway is a molecular network regulating the process from the completion of one cell division to the next. m6A modification is involved in cell cycle control [12]. During the G2 phase of adipogenesis, knockdown of FTO has been shown to delay cell cycle progression [13]. In MM, this pathway is dysregulated due to genetic and epigenetic alterations [14]. Notably, FTO and the cell cycle pathway share common pathogenic mechanisms in MM. From a genetic-microenvironment synergy perspective, IL-6/IGF-1 in the bone marrow microenvironment activates the PI3K-Akt-GSK3β axis to stabilize Cyclin D1, while chromosomal abnormalities such as t(11;14) also drive Cyclin D1 overexpression. These mechanisms synergize with FTO’s epigenetic regulation to amplify cell cycle signaling [15]. In summary, FTO, as an m6A demethylase, acts as a “cell cycle modulator” in the pathogenesis of MM. By concurrently weakening p21 and stabilizing MCL-1 through its demethylation, FTO drives dysregulated cell cycle progression in MM cells. Thus, targeting FTO may represent a therapeutic strategy to suppress MM cell proliferation.

Our analysis demonstrates that both CHRM3 and FTO were synchronously and significantly upregulated in MM tissues, suggesting a critical role of the FTO–CHRM3 axis in the pathogenesis of the disease. CHRM3 (Cholinergic Muscarinic Receptor 3), an M3 subtype of muscarinic acetylcholine receptors, belongs to the G protein-coupled receptor family and primarily mediates parasympathetic signaling of acetylcholine (ACh). Its activation triggers downstream signaling cascades via Gαq/11 proteins. Existing studies indicate that CHRM3 exhibits oncogenic functions in various solid tumors [16, 17]. Although MM is a hematologic malignancy with a pathogenesis distinct from that of solid tumors, CHRM3 has been shown to promote proliferation and inhibit apoptosis in MM through autocrine ACh–CHRM3 signaling. Furthermore, it synergizes with bone marrow stromal remodeling to facilitate drug resistance and homing [18]. Its high expression is significantly associated with shorter patient survival, supporting its potential as a novel prognostic biomarker and therapeutic target.

Based on GSEA results, this study further identifies the calcium signaling pathway involving CHRM3 as an m6A methylation-related core pathway. The calcium signaling pathway is a crucial intracellular signal transduction mechanism that modulates various physiological processes such as proliferation, differentiation, apoptosis, and migration by regulating changes in cytosolic Ca²⁺ concentration. In MM, overexpression of FTO, mediated through stabilization of CHRM3 mRNA, results in sustained high expression of CHRM3 [19], which subsequently triggers massive Ca²⁺ release from intracellular stores via activation of the Gq/11–PLC–IP₃ axis [20]. Aberrantly activated calcium signaling subsequently drives myeloma cell proliferation, survival, migration, and drug resistance through downstream pathways such as CaMKII–NF-κB/NFAT [20, 21], while also exacerbating bone destruction by enhancing osteoclast activity and suppressing osteogenesis. Thus, it serves as a key driver in the initiation and progression of MM. The FTO–CHRM3–Calcium Signaling Pathway is proposed to contribute to the pathogenesis and progression of MM via aberrant activation of calcium signaling (Figure S7).

In conclusion, we identify a novel FTO-CHRM3 regulatory axis as a driver of MM progression. Our work demonstrates that FTO-mediated epigenetic dysregulation of CHRM3 disrupts calcium signaling, revealing a previously unrecognized mechanism in MM oncogenesis and establishing this axis as a promising target for therapeutic intervention. However, this study has limitations, such as the absence of experimental validation (e.g., gene knockout/pathway assays), thereby limiting the conclusions to correlative rather than causal relationships. Due to public database limitations, this study cannot fully correct clinical confounding factors, control for tumor purity and primary cancer content in bulk bone marrow samples, or eliminate batch/platform effects across multi-platform datasets. The association between CHRM3 and methylation was not further investigated in this study, and the specific regulatory effect and downstream signaling pathway of the FTO-CHRM3 axis were only inferred based on correlation analyses, with their exact functional mode remaining to be verified by subsequent experimental studies. Additionally, the immunohistochemical verification used a relatively small sample size of 3:3, and further studies with a larger sample size are needed to confirm these results. In the exploratory single-cell analysis, no significant differential expression was observed for FTO and CHRM3 in cell types. This result might be influenced by factors such as sample heterogeneity, technical detection thresholds, and the accuracy of cell annotation. In the future, we will further verify this through increasing the sample size, optimizing the experimental and analytical strategies.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ACh

Acetylcholine

BP

Biological Process

CAR-T

Chimeric Antigen Receptor T-cell

CC

Cellular Component

C-index

Concordance Index

DEGs

Differentially Expressed Genes

Enet

Elastic Network

FC

Fold Change

FTO

Fat Mass and Obesity-Associated

GBM

Gradient Boosting Machine

GO

Gene Ontology

GSEA

Gene Set Enrichment Analysis

GSVA

Gene Set Variation Analysis

HPA

Human Protein Atlas

HR

Hazard Ratio

IC₅₀

Half Maximal Inhibitory Concentration

KM

Kaplan-Meier

LASSO

Least Absolute Shrinkage and Selection Operator

m6A

N6-methyladenosine

MF

Molecular Function

MM

Multiple Myeloma

NK cells

Natural Killer cells

PH

Proportional Hazards

plsRcox

Partial Least Squares Cox Regression

PPI

Protein-Protein Interaction

RSF

Random Survival Forest

scRNA-seq

Single-cell RNA Sequencing

SuperPC

Supervised Principal Components

SVM-RFE

Support Vector Machine-Recursive Feature Elimination

TIDE

Tumor Immune Dysfunction and Exclusion

Tregs

Regulatory T cells

tSNE

t-distributed Stochastic Neighbor Embedding

vst

Variance Stabilizing Transformation

Authors’ contributions

BL, TB, and X-FD are the principal investigator and conducted statistical analysis and drafted the article. CL, JT, and T-TS performed data management and bioinformatics analysis. BL, TB, X-FD, MX, CL, JT, and T-TS edited and revised the article. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by Sanming Project of Medicine in Shenzhen (grant number: SZSM202411022) and Shenzhen Science and Technology Innovation Commission (grant number: JCYJ20230807110659001).

Data availability

The datasets (GSE118985, GSE24870, GSE24080, and GSE124310) were collected from Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/), and the MMRF-COMMPASS dataset was downloaded from UCSC Xena platform (https://xenabrowser.net/).

Declarations

Ethics approval and consent to participate

The studies involving human participants were reviewed and approved by The Seventh Affiliated Hospital, Sun Yat-Sen University (Approval No. KY-2025-438-01). Written informed consent to participate in this study was provided by the participants. The study protocols adhered strictly to the principles outlined in the Helsinki Declaration.

Competing interests

The authors declare no competing interests.