A bioinformatics and experimental approach identifies glycoprotein-based diagnostic and prognostic biomarkers for colon adenocarcinoma Authors: Haiyun Liu, Linjing Wan, Quangang Fang Published: 2026 ABSTRACT AbstractColon adenocarcinoma (COAD) remains a leading cause of cancer-related mortality worldwide, partly due to the lack of robust biomarkers for early diagnosis and accurate prognosis. Glycoproteins play critical roles in tumorigenesis and may serve as promising sources of biomarkers. This study aimed to identify glycoprotein-related candidate diagnostic and prognostic biomarkers for COAD through integrated bioinformatics approaches and experimental validation. Gene expression profiles and clinical information of COAD patients were obtained from The Cancer Genome Atlas database (TCGA). The expression data for genes associated with glycoprotein were retrieved from the UniProt database. Furthermore, we assessed the mRNA expression levels of the glycoprotein FJX1 in COAD and rectal adenocarcinoma tissues through in situ hybridization (ISH) staining using tissue microarrays. A total of 228 glycoprotein-related differentially expressed genes (DEGs) were identified, enriched in the extracellular matrix organization and signaling pathways such as PI3K-Akt and cAMP. Protein-protein interaction (PPI) network analysis revealed 10 hub genes (LIFR, CNTFR, LIF, CSF2, IL1A, CSF3, F2, FGA, FGFR2, INHBA). Survival screening and multivariate Cox regression identified FJX1 as an independent prognostic factor for overall survival after adjusting for age and stage. FJX1 mRNA was significantly overexpressed in COAD tissues compared to normal (p < 0.001), and high FJX1 expression correlated with advanced T stage, M stage, and pathological stage. ISH confirmed elevated FJX1 mRNA in tumors. Immune infiltration analysis further revealed that high FJX1 expression was associated with increased infiltration of M0 macrophages and neutrophils, and decreased resting memory CD4+ T cells, suggesting a potential role in shaping the immunosuppressive tumor microenvironment. GSEA revealed significant enrichment of MAPK signaling in high-FJX1 tumors. In conclusion, this study identified 10 glycoprotein-related hub genes as candidate diagnostic biomarkers warranting further validation and established FJX1 as an independent prognostic biomarker for COAD. FJX1 overexpression is associated with tumor progression and may be linked to MAPK signaling as well as immune modulation. These findings provide a foundation for glycoprotein-based biomarker development in COAD.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-026-52427-5.Keywords: Colon adenocarcinoma, Glycoprotein, Biomarker, FJX1, Prognosis, MAPK pathway, Immune infiltrationSubject terms: Biomarkers, Cancer, Computational biology and bioinformatics, Oncology Introduction Colon adenocarcinoma (COAD) is one of the most common malignancies worldwide1, with increasing incidence and high mortality rates. These rates are largely attributable to late diagnosis and the lack of reliable biomarkers for prognostic stratification and treatment guidance2,3. Currently used tumor markers, such as carcinoembryonic antigen (CEA) and carbohydrate antigen 19 − 9 (CA19-9), have limited sensitivity and specificity, particularly when it comes to early detection and monitoring of disease progression4,5. Therefore, the identification of novel, robust biomarkers with diagnostic and prognostic value is urgently needed to improve clinical management of COAD. Glycoproteins are proteins that are modified by the covalent attachment of carbohydrate moieties6, playing essential roles in various biological processes, including cell adhesion, signaling, and immune recognition7–9. Aberrant glycosylation is a hallmark of cancer, and many glycoproteins are differentially expressed in tumor tissues compared to their normal counterparts10,11. Compared with conventional protein biomarkers, glycoproteins offer the advantage of reflecting both protein expression and post-translational modifications, potentially providing more specific insights into cancer-associated alterations12. Moreover, glycoproteins have been implicated in tumor progression, metastasis, and therapeutic resistance13–16. These characteristics make glycoproteins attractive candidates for biomarker discovery. Recent advances in high-throughput technologies and bioinformatics have enabled the systematic exploration of glycoprotein-related genes in cancer10,17. FJX1 (four-jointed box protein 1) is a transmembrane glycoprotein originally identified as a regulator of planar cell polarity in Drosophila18. Emerging evidence suggests its involvement in human cancers, including nasopharyngeal carcinoma19, gastric cancer20, and colorectal cancer21. In COAD, FJX1 has been reported to promote cell migration and invasion in vitro22. A recent study showed that metformin suppresses COAD viability by modulating FJX1 through miR-1306-3p23. However, the clinical significance, prognostic value, and underlying mechanisms of FJX1 in COAD remain poorly understood. In this study, we employed an integrated bioinformatics approach combining TCGA data analysis, PPI network construction, and survival modeling to identify glycoprotein-related genes with diagnostic and prognostic potential in COAD. We identified a 10-gene hub network with diagnostic implications as candidate biomarkers and further focused on FJX1, identified as a key prognostic biomarker. Using an independent clinical cohort, we validated the overexpression of FJX1 at the mRNA level and demonstrated its association with clinicopathological features. Our findings provide a resource for glycoprotein-based biomarker discovery and highlight FJX1 as a promising candidate for COAD. Materials and methods Data acquisition and processing RNA-seq data (level 3 HTSeq-FPKM) and corresponding clinical information for COAD were downloaded from The Cancer Genome Atlas (TCGA) database (https://portal.gdc.cancer.gov/). The dataset included 401 tumor samples and 39 adjacent normal samples. FPKM values were log2-transformed after adding a pseudo-count of 1. Genes with low expression (mean log2(FPKM + 1) < 1 across all samples) were filtered out. No additional batch correction was applied, as the data were generated by a single platform24. The overall study design is illustrated in Fig. 1, and detailed clinical characteristics of the COAD patients are provided in Supplementary Table S1. Methodological workflow. A schematic illustration of the overall study design and analytical workflow. Identification of glycoprotein-related genes A list of glycoprotein-related genes was compiled from the UniProt database (https://www.uniprot.org/) using the following criteria: reviewed (Swiss-Prot) entries, organism “Homo sapiens”, and keyword “Glycoprotein” (KW-0325). This search yielded 1,998 genes with experimentally verified or curated glycoprotein annotations25. The list was downloaded in January 2024. Differential expression analysis Differential expression analysis between tumor (n = 401) and normal (n = 39) samples was performed using the limma package (v4.2.3) in R26. Genes with |log2 fold change (FC)| > 2 and false discovery rate (FDR) < 0.05 were considered significantly differentially expressed (DEGs). Functional enrichment analysis Overlapping glycoprotein-related DEGs were subjected to Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses using the clusterProfiler package (v4.6.0) in R27. Enriched terms with FDR q-value < 0.05 were considered statistically significant. Protein-protein interaction network and hub gene selection The overlapping glycoprotein-related DEGs were submitted to the STRING database (v11.5, https://cn.string-db.org/) to construct a PPI network28. Interactions with a combined score > 0.9 (high confidence) were retained, and disconnected nodes were hidden. The network was exported to Cytoscape (v3.9.1) for visualization and topological analysis. Hub genes were identified using the CytoHubba plugin29 based on degree centrality, with the top 10 genes selected as hub genes. Survival analysis and prognostic model Univariate Cox proportional hazards regression was performed to assess the association between each glycoprotein-related DEG and overall survival (OS) in TCGA-COAD patients using the survival package (v3.5-5) in R. Genes with p < 0.05 were considered candidate prognostic genes. To control for multiple testing, we applied the Benjamini-Hochberg FDR correction; genes with FDR < 0.2 were retained. Multivariate Cox regression with stepwise selection was then used to identify independent prognostic factors, adjusting for age, gender, and TNM stage. The proportional hazards assumption was tested using Schoenfeld residuals. The optimal cut-off for FJX1 expression was determined by X-tile30 and confirmed by maximally selected rank statistics. Kaplan-Meier survival curves were compared by log-rank test. Additionally, time-dependent receiver operating characteristic (ROC) curves were generated using the timeROC R package to evaluate the prognostic accuracy of FJX1 at 1, 3, and 5 years. Immune infiltration analysis To explore the relationship between FJX1 expression and the tumor immune microenvironment in COAD, we employed two computational approaches. First, the CIBERSORT algorithm was applied to the TCGA-COAD RNA-seq data to estimate the relative proportions of 22 distinct immune cell types in each tumor sample. Differences in immune cell infiltration between high- and low-FJX1 expression groups (stratified by median expression) were assessed using the Wilcoxon rank-sum test and visualized with violin plots. Second, single-sample Gene Set Enrichment Analysis (ssGSEA) was performed using the GSVA R package to calculate enrichment scores for 28 immune cell signatures. Spearman correlation analysis was then used to evaluate the association between FJX1 mRNA levels and ssGSEA-derived immune cell enrichment scores. All analyses were conducted in R (v4.2.1). Tissue microarrays (TMA) construction and in situ hybridization (ISH) TMAs were constructed from 84 formalin-fixed, paraffin-embedded (FFPE) colorectal cancer (CRC) tissue sections following the protocol described by Nocito et al.31. Tissues and corresponding clinical data (Supplementary Table S2) were collected from CRC patients at Jiangxi Provincial People’s Hospital between October 2023 and June 2025. Patient Inclusion Criteria: (1) had a primary diagnosis of CRC confirmed by histopathology and had resided in Nanchang for over 5 years; (2) were at average risk for CRC (no prior CRC diagnosis or known high-risk hereditary syndromes); (3) were undergoing a standard colonoscopy following a positive Fecal Immunochemical Test (FIT). Patient Exclusion Criteria: (1) a history of any prior malignancy (except non-melanoma skin cancer); (2) a history of inflammatory bowel disease; (3) previous surgical colon resection; (4) presence of urgent clinical symptoms (e.g., significant rectal bleeding, iron-deficiency anemia, unexplained weight loss); (5) severe comorbidities compromising follow-up; (6) a colonoscopy, sigmoidoscopy, or CT colonography within the 5 years prior to enrollment. In situ hybridization was carried out as previously outlined32. Briefly, TMA sections were dewaxed, rehydrated, and subjected to antigen retrieval in boiling buffer. After digestion with Proteinase K and blocking of endogenous peroxidase, slides were incubated with a pre-hybridization solution. Hybridization was performed using a mixture of FJX1-targeting probes (Table 1) (designed by Servicebio, Wuhan). Post-hybridization washes were conducted with SSC buffers. Hybridized probes were detected using an mouse anti-digoxin labeled peroxidase (anti-DIG-HRP) conjugate and visualized with DAB substrate. Finally, sections were counterstained with hematoxylin, dehydrated, and mounted. Table 1A set of in situ hybridization probes designed to specifically detect FJX1 gene sequences.Target geneSequenceFJX1ACTGCAACAGCGGCTCGTTATACTTCAAAATGTGCTTGGCGAGGAAGTCGAGTTGGCCGTCAGGTAGTCGAAAAGCCGCCTCATTGTCCAGAAAGACCAG A set of in situ hybridization probes designed to specifically detect FJX1 gene sequences. Evaluation of ISH staining The staining results were quantitatively assessed using Aipathwell software (Servicebio). An H-Score was calculated for each sample based on staining intensity and the percentage of positive cells, using the formula: H-Score = (percentage of weakly stained cells × 1) + (percentage of moderately stained cells × 2) + (percentage of strongly stained cells × 3). This yields a score ranging from 0 to 300. The positive rate was defined as the total percentage of cells exhibiting any positive staining. Prediction of drug, microRNA, and transcription factor interactions Potential interactions with the identified hub genes were explored using several databases. The DGIdb database (https://dgidb.genome.wustl.edu/) was queried to predict small-molecule drugs that interact with the hub genes. The miRNet database (https://www.mirnet.ca/) was used to identify microRNAs (miRNAs) potentially targeting the hub genes. Furthermore, the same miRNet database was employed to predict transcription factors (TFs) that might regulate the hub genes. Gene set enrichment analysis (GSEA) GSEA (v4.3.2) was performed to identify KEGG pathways differentially enriched between high and low FJX1 expression groups in TCGA-COAD33. The expression cutoff was based on the median FPKM. Gene sets were obtained from the Molecular Signatures Database (MSigDB, v2023.2, c2.cp.kegg_legacy.v7.5.1.symbols.gmt). Enrichment was assessed using 1000 permutations, with significance defined as nominal p < 0.05 and FDR q < 0.25. Statistical analysis All statistical analyses were performed using R software (v4.2.1) and GraphPad Prism (v9.0). Differences between two groups were assessed by Student’s t-test or Mann-Whitney U test, as appropriate. Paired samples were compared using paired t-test or Wilcoxon signed-rank test. Categorical variables were analyzed by chi-square or Fisher’s exact test. Survival curves were estimated by Kaplan-Meier method and compared by log-rank test. Multivariate Cox regression was used to identify independent prognostic factors. A two-sided p < 0.05 was considered statistically significant, with FDR correction applied for multiple comparisons where indicated. Results Identification of glycoprotein-related DEGs in COAD After preprocessing, the TCGA-COAD expression matrix contained 19,744 genes across 401 tumors and 39 normals. Differential expression analysis identified 2,637 DEGs (|log2FC| > 2, FDR < 0.05), comprising 1,988 upregulated and 649 downregulated genes in tumors (Fig. 2A; detailed in Supplementary Table S3). Intersection with the glycoprotein-related gene list (n = 1,998) (Supplementary Table S4) yielded 228 overlapping genes (Fig. 2B), designated as glycoprotein-related DEGs. Identification of candidate hub genes. (A) Volcano plot of differentially expressed genes (DEGs) in colon adenocarcinoma (COAD) compared to adjacent normal tissue. Significantly up-regulated genes (log2FC > 2 and FDR < 0.05) and down-regulated genes (log2FC < -2 and FDR < 0.05) are highlighted in red and green, respectively. (B) A Venn diagram illustrating the screening of candidate hub genes, showing the overlap between DEGs in COAD and glycoprotein-related genes. Functional enrichment analysis of glycoprotein-related DEGs To understand the biological roles of the 228 glycoprotein-related DEGs, we performed GO and KEGG enrichment analyses. GO analysis categorized the genes into three domains: (1) Biological Process (BP): The DEGs were significantly enriched in processes including wound healing, extracellular matrix (ECM) organization, extracellular structure organization, and external encapsulating structure organization; (2) Cellular Component (CC): They were primarily localized to the external side of the plasma membrane, collagen-containing extracellular matrix, anchored component of the membrane, and the apical part of the cell; (3) Molecular Function (MF): The main functions included receptor ligand activity and signaling receptor activator activity (Fig. 3A). Functional enrichment analysis of overlapping glycoprotein-related genes. (A) Gene Ontology (GO) enrichment analysis. (B) Kyoto Encyclopedia of Genes and Genomes (KEGG)34–36 pathway enrichment analysis. For previous uses, the Kanehisa laboratory has kindly provided permission. KEGG pathway analysis indicated that these genes were predominantly involved in neuroactive ligand-receptor interaction, the phosphatidylinositol-3-kinase/AKT (PI3K-Akt) signaling pathway, and cytokine-cytokine receptor interaction (FDR q < 0.05) (Fig. 3B). PPI network and hub gene identification The PPI network constructed from 228 glycoprotein-related DEGs with a high confidence score (> 0.9) contained 221 nodes and 40 edges, with many disconnected nodes hidden (Fig. 4A). The network was relatively sparse, consisting mainly of small clusters. Topological analysis using CytoHubba identified the top 10 hub genes based on degree centrality: LIFR, CNTFR, LIF, CSF2, IL1A, CSF3, F2, FGA, FGFR2, and INHBA (Fig. 4B). These genes represent highly connected nodes within their respective subnetworks and are proposed as candidate diagnostic biomarkers. Identification of glycoprotein-related hub DEGs in COAD. (A) Protein-protein interaction (PPI) network constructed from the overlapping genes. (B) Hub genes identified from the PPI network using Cytoscape. Prediction of regulatory elements and drugs for hub genes We next explored potential regulators and therapeutic agents targeting the 10 hub genes. Drug-Gene Interactions: The DGIdb database predicted 92 small-molecule drugs that could interact with the hub genes. For instance, six drugs (e.g., AMG-108, Bermekimab) were linked to IL1A, and four drugs (e.g., Benegrastim, Eflapegrastim) were associated with CSF3. Other specific interactions included ciliary neurotrophic factor for CNTFR, Sotatercert for INHBA, and Emfilermin for LIFR (Fig. 5A); MicroRNA and Transcription Factor (TF) Networks: Using the miRNet database, we constructed regulatory networks, identifying 29 microRNAs (Fig. 5B) and 11 transcription factors (Fig. 5C) that potentially interact with the hub genes. Regulatory networks of hub genes, predicted regulators, and therapeutic agents targeting the identified hub genes. (A) Small-molecule drugs potentially modulating hub gene expression or function. (B) MicroRNAs predicted to target the hub genes. (C) Transcription factors predicted to regulate hub gene expression. Prognostic value of glycoprotein-related genes and identification of FJX1 Survival analysis of the 228 glycoprotein-related DEGs revealed 17 genes significantly correlated with OS in COAD patients (Table 2). Univariate Cox regression analysis on these 17 genes identified five—GABRD, ERFE, LDLRAD3, FJX1, and GRIK5—as being significantly associated with OS (Fig. 6A). Table 2A list of genes significantly associated with colon adenocarcinoma survival among 228 overlapping differentially expressed genes (DEGs).GeneKaplan-MeierChloride channel accessory 4 (CLCA4)0.0147HEPACAM family member 2 (HEPACAM2)0.0151Gamma-aminobutyric acid type A receptor subunit delta (GABRD)0.0045Integrin subunit beta like 1 (ITGBL1)0.0066Immunoglobulin heavy constant alpha 2 (IGHA2)0.0182Chloride channel accessory 1 (CLCA1)0.0005Erythroferrone (ERFE)0.0026Low density lipoprotein receptor class A domain containing 3 (LDLRAD3)0.0109Fc gamma binding protein (FCGBP)0.0240Four-jointed box protein 1 (FJX1)0.0002Collagen and calcium binding EGF domains 1 (CCBE1)0.0042Integral membrane protein 2 C (ITM2C)0.0370Glutamate ionotropic receptor kainate type subunit 5 (GRIK5)0.0024Cartilage intermediate layer protein 2 (CILP2)0.0095Interleukin 1 alpha (IL1A)0.0168Intelectin 1 (ITLN1)0.0093Gliomedin (GLDN)0.0087 A list of genes significantly associated with colon adenocarcinoma survival among 228 overlapping differentially expressed genes (DEGs). Prognostic analysis of glycoprotein-related genes in COAD. (A) Univariate Cox regression analysis identified five glycoprotein-related DEGs that are significantly associated with overall survival (OS). (B-C) Forest plots from univariate (B) and multivariate (C) Cox regression analyses demonstrate the relationship between FJX1 expression and OS in COAD patients. Univariate and multivariate Cox regression incorporating these five genes along with clinical parameters (age, gender, stage, T, N, M) revealed that FJX1 was an independent prognostic factor for OS after stepwise selection (Fig. 6B-C), while the other four genes were not retained. The proportional hazards assumption was satisfied (p > 0.05 for all covariates). Kaplan-Meier analysis using the optimal cut-off (X-tile) showed that patients with high FJX1 expression had significantly poorer OS compared to those with low expression (log-rank p < 0.001; Fig. 7A). Additionally, time-dependent ROC curve analysis demonstrated that FJX1 expression had prognostic value, with AUC values of 0.637, 0.646, and 0.695 for predicting 1-, 3-, and 5-year survival, respectively (Fig. 7B). Survival and diagnostic analysis of FJX1. (A) Kaplan-Meier survival curve showing the correlation between FJX1 expression levels and OS in patients with COAD. (B) Receiver operating characteristic (ROC) curve evaluating the diagnostic potential of FJX1 for COAD. FJX1 expression and clinicopathological correlations In TCGA data, FJX1 mRNA expression was significantly higher in tumor tissues than in normal tissues (p < 0.001; Fig. 8A). High FJX1 expression (above median) correlated with advanced pathological stage (p = 0.039), T stage (p = 0.047), and M stage (p = 0.036), but not with age, gender, or N stage (Fig. 8B-D). These observations were supported by logistic regression analysis, which confirmed a significant association between FJX1 expression and both advanced pathological stage and M classification (Table 3). FJX1 Expression and its Correlation with Clinicopathological Features. (A) Comparison of the FJX1 expression levels between COAD tumors and matched normal tissues (p < 0.001). (B–D) The FJX1 expression levels stratified by pathological stage (B), T stage (C), and M stage (D). Table 3Association between FJX1 expressiona and clinical pathological characteristics (logistic regression).Clinical characteristicsTotal (N)Odds ratio in FJX1 expressionp-ValuePathological stage Stage II vs. Stage I2151.4074 (0.7842–2.5506)0.2547 Stage III vs. Stage I1671.2510 (0.6690–2.3570)0.4846 Stage IV vs. Stage I1192.5926 (1.2422–5.5398)0.0122M classification M1 vs. M03372.0776 (1.1459–3.8688)0.0179a Categorical dependent variable, greater or less than the median expression level. Association between FJX1 expressiona and clinical pathological characteristics (logistic regression). a Categorical dependent variable, greater or less than the median expression level. Validation of FJX1 overexpression in clinical specimens To further validate the potential of FJX1 as a biomarker for COAD, we collected 42 clinical specimens comprising CRC tumor tissues and their corresponding adjacent normal tissues, which included 19 cases of COAD and 20 cases of rectal carcinoma (Rectal CA). These specimens were subsequently organized into tissue microarrays. As illustrated in the scatter plots and paired expression graphs (Fig. 9A-F and Supplementary Table S5), ISH revealed that FJX1 mRNA expression levels were significantly elevated in both COAD and Rectal CA, as well as in the overall CRC, when compared to adjacent normal tissues. Furthermore, ISH staining corroborated these findings, demonstrating a marked overexpression of FJX1 mRNA in COAD and Rectal CA relative to adjacent normal tissues (Fig. 9G-H and Supplementary Figure S1). In situ hybridization staining was employed to assess the expression levels of FJX1 mRNA in both colorectal adenocarcinoma (COAD) and rectal carcinoma (Rectal CA). Panels (A-F) present scatter plots (A, C, and E) and paired plots (B, D, and F) illustrating the expression patterns of FJX1 mRNA in colorectal cancer (CRC) (A, B), COAD (C, D), and rectal CA (E, F). Panels (G–H) display representative in situ hybridization images with DAB staining, depicting FJX1 mRNA expression in COAD (G) or rectal CA (H) alongside adjacent normal tissues. Identification of FJX1-associated signaling pathways via GSEA To investigate the potential mechanisms by which FJX1 influences COAD progression, we performed GSEA. This identified several signaling pathways significantly enriched in samples with high FJX1 expression (NOM p-val < 0.05, FDR q-val < 0.05). The top enriched pathways included MAPK signaling pathway, focal adhesion, ECM-receptor interaction, and cell adhesion molecules (CAMs) (Fig. 10A-B; detailed in Tables 4 and 5). Gene Set Enrichment Analysis (GSEA). (A-B) Enrichment plots from GSEA identifying KEGG34–36 pathways significantly enriched in phenotypes with high (A) and low (B) FJX1 expression. For previous uses, the Kanehisa laboratory has kindly provided permission. Table 4Gene sets enriched in high FJX1 expression phenotype.MSigDB collectionGene set nameNESNOM p-valFDR q-valc2.cp.kegg_legacy.v2023.2.Hs.symbols.gmtKEGG_FOCAL_ADHESION3.200.0000.000KEGG_ECM_RECEPTOR_INTERACTION3.160.0000.000KEGG_PATHWAYS_IN_CANCER2.980.0000.000KEGG_CHEMOKINE_SIGNALING_PATHWAY2.960.0000.000KEGG_CELL_ADHESION_MOLECULES_CAMS2.940.0000.000KEGG_CYTOKINE_CYTOKINE_RECEPTOR_INTERACTION2.900.0000.000KEGG_LEUKOCYTE_TRANSENDOTHELIAL_MIGRATION2.750.0000.000KEGG_BASAL_CELL_CARCINOMA2.690.0000.000KEGG_MAPK_SIGNALING_PATHWAY2.650.0000.000KEGG_WNT_SIGNALING_PATHWAY2.430.0000.000NES: normalized enrichment score; NOM: nominal; FDR: false discovery rate. Gene sets with NOM p-val < 0.05 and FDR q-val < 0.05 are considered as significant. Gene sets enriched in high FJX1 expression phenotype. c2.cp.kegg_legacy.v2023. 2.Hs.symbols.gmt NES: normalized enrichment score; NOM: nominal; FDR: false discovery rate. Gene sets with NOM p-val < 0.05 and FDR q-val < 0.05 are considered as significant. Table 5Gene sets enriched in low FJX1 expression phenotype.MSigDB collectionGene set nameNESNOM p-valFDR q-valc2.cp.kegg_legacy.v2023.2.Hs.symbols.gmtKEGG_OLFACTORY_TRANSDUCTION– 2.250.0000.000KEGG_ASCORBATE_AND_ALDARATE_METABOLISM– 1.930.0020.003KEGG_VALINE_LEUCINE_AND_ISOLEUCINE_DEGRADATION– 1.840.0060.007KEGG_DRUG_METABOLISM_CYTOCHROME_P450– 1.770.0020.011KEGG_METABOLISM_OF_XENOBIOTICS_BY_CYTOCHROME_P450– 1.700.0180.303KEGG_PROPANOATE_METABOLISM– 1.690.0060.017KEGG_PENTOSE_AND_GLUCURONATE_INTERCONVERSIONS– 1.680.0080.016KEGG_STEROID_HORMONE_BIOSYNTHESIS– 1.640.0060.020KEGG_DRUG_METABOLISM_OTHER_ENZYMES– 1.640.0080.019KEGG_RETINOL_METABOLISM– 1.580.0080.029NES: normalized enrichment score; NOM: nominal; FDR: false discovery rate. Gene sets with NOM p-val < 0.05 and FDR q-val < 0.05 are considered as significant. Gene sets enriched in low FJX1 expression phenotype. c2.cp.kegg_legacy.v2023. 2.Hs.symbols.gmt NES: normalized enrichment score; NOM: nominal; FDR: false discovery rate. Gene sets with NOM p-val < 0.05 and FDR q-val < 0.05 are considered as significant. Association of FJX1 expression with tumor-infiltrating immune cells Given the pivotal role of immune infiltration in tumor progression, we further investigated the relationship between FJX1 expression and the tumor immune microenvironment in COAD. Using the CIBERSORT algorithm, we first assessed the relative abundance of 22 immune cell subsets in TCGA-COAD samples (Fig. 11A). Correlation analysis among these immune cell types revealed a strong positive association between resting mast cells and M2 macrophages, as well as between resting NK cells and activated memory CD4+ T cells. Conversely, a negative correlation was observed between M0 macrophages and plasma cells, and between activated mast cells and resting mast cells (Fig. 11B). Immune infiltration landscape of COAD and its association with FJX1 expression. (A) The relative proportions of 22 immune cell types in COAD tumors were estimated using the CIBERSORT deconvolution algorithm. Each stacked bar represents the immune cell composition of an individual patient, with the corresponding cell subsets color-coded in the legend on the right. (B) Pairwise correlation analysis among the 22 immune cell fractions. Positive correlations are shown in red, and negative correlations are shown in blue. (C) Comparison of immune cell subtype abundances between the FJX1-low expression group (green) and the FJX1-high expression group (red). (D) Heatmap illustrating the Spearman correlation coefficients between FJX1 mRNA expression and immune cell infiltration scores determined by ssGSEA. We then stratified COAD patients into high- and low-FJX1 expression groups based on median FJX1 mRNA levels. Violin plot analysis demonstrated that the high-FJX1 group exhibited significantly increased infiltration of M0 macrophages (p < 0.05) and neutrophils (p < 0.05), whereas infiltration levels of resting memory CD4⁺ T cells (p < 0.05) and monocytes (p < 0.05) were significantly reduced compared to the low-FJX1 group (Fig. 11C). Complementary analysis using ssGSEA further explored correlations between FJX1 expression and 28 immune cell signatures. As shown in Fig. 11D, Spearman correlation analysis revealed significant positive correlations between FJX1 expression and multiple immune cell types, including plasmacytoid dendritic cells, regulatory T cells, gamma delta T cells, type 1 T helper cells, central memory CD8+ T cells, natural killer cells, macrophages, effector memory CD4+ T cells, immature dendritic cells, mast cells, natural killer T cells, myeloid-derived suppressor cells (MDSCs), T follicular helper cells, activated dendritic cells, central memory CD4+ T cells, CD56bright natural killer cells, and neutrophils. These findings collectively suggest that FJX1 overexpression is associated with a distinct and potentially immunosuppressive immune landscape in COAD. Discussion The high mortality of COAD is largely attributable to late diagnosis and a lack of robust biomarkers to guide prognostic stratification and treatment decisions37. In recent years, researchers have increasingly focused on investigating crucial proteins and genes implicated in tumor occurrence, progression, and metastasis. Glycoproteins are proteins that bind with carbohydrate moieties6. Studies have revealed that glycoproteins exhibit differential expression in tumor and normal tissues and contribute to tumor resistance38,39. Furthermore, the contents of glycoprotein are altered during tumor development40–42. Therefore, in this study, we systematically identified glycoprotein-related genes with diagnostic and prognostic potential in COAD through integrated bioinformatics and experimental validation. Our main findings are: (i) identification of a 10-gene hub network (LIFR, CNTFR, LIF, CSF2, IL1A, CSF3, F2, FGA, FGFR2, INHBA) as candidate diagnostic biomarkers; (ii) establishment of FJX1 as an independent prognostic biomarker associated with poor survival; (iii) validation of FJX1 mRNA overexpression in an independent clinical cohort; and (iv) association of FJX1 with MAPK signaling pathway by GSEA; and (v) demonstration that high FJX1 expression correlates with distinct tumor-infiltrating immune cell profiles. The ten hub genes identified through our PPI network analysis represent a promising diagnostic signature for COAD that warrants further validation. The literature robustly supports the involvement of these genes in tumorigenesis across various cancers, lending credence to their selection. For instance, the LIF/LIFR signaling axis is known to remodel the tumor immune microenvironment in solid malignancies43, while CNTFR dysregulation has been documented in oral squamous cell carcinoma44. CSF2 is associated with chemoresistance in colorectal cancer45, and the coagulation factor F2 (prothrombin) is a key player in cancer-associated thrombosis46. Furthermore, IL1A is a well-characterized regulator of tumor progression and angiogenesis47. The roles of FGA48,49, FGFR250, CSF351, and INHBA52 in promoting cancer cell proliferation, survival, and metastasis have also been established in gastrointestinal and other cancers. The construction of associated drug-miRNA-TF networks for these hub genes provides a valuable resource for hypothesizing regulatory mechanisms and potential therapeutic strategies, many elements of which have been previously validated in other pathological contexts53,54. Collectively, these ten glycoprotein-related genes form a compelling basis for developing a molecular diagnostic panel for COAD, though we acknowledge that formal assessment of diagnostic performance (e.g., ROC analysis, sensitivity, and specificity in independent cohorts) is required before clinical application. Beyond its diagnostic implications, our study provides compelling clinical evidence establishing FJX1 as a central prognostic determinant in COAD. Although recent studies have reported differential expression and significant biological roles of FJX1 in various cancers, including nasopharyngeal carcinoma19, colorectal cancer21, and gastric cancer20, and prior in vitro work has suggested that FJX1 promotes migration and invasion in COAD55. Kim et al.23 have demonstrated that metformin modulates FJX1 through the upregulation of Hsa-miR-1306-3p, thereby suppressing COAD viability in vitro. Nonetheless, the clinical significance of these findings remains inadequately elucidated. Our findings bridge this gap by demonstrating that FJX1 expression is not only significantly elevated in COAD tissues but also positively correlated with advanced disease stage, including T stage and metastasis (M stage). Crucially, multivariate Cox regression analysis confirmed FJX1 as an independent risk factor for overall survival, with Kaplan-Meier analysis revealing a stark survival disadvantage for patients with high FJX1 expression. The robust AUC values from our time-dependent ROC curve analysis in the TCGA cohort further underscore its potential reliability in predicting patient outcomes at 1, 3, and 5 years. However, we acknowledge that our clinical validation cohort lacked survival follow-up data, precluding independent prognostic confirmation. Prospective studies with complete follow-up are essential to fully establish FJX1 as a clinically actionable prognostic biomarker. To elucidate the potential mechanistic role of FJX1, we performed GSEA, which revealed a significant enrichment of the MAPK signaling pathway in tumors with high FJX1 expression. The MAPK pathway is a well-established driver of COAD pathogenesis, regulating critical processes such as cell proliferation, invasion, and survival56. This finding offers a plausible molecular explanation for FJX1’s oncogenic effects and is consistent with a prior report in gastric cancer, where FJX1 was shown to facilitate tumor development via MAPK signaling57. Therefore, our data suggest that FJX1 may contribute to COAD aggressiveness through the activation of this key oncogenic pathway. Notably, this finding aligns with reports on other glycoproteins in CRC. For instance, Syndecan-2 (SDC2), a transmembrane heparan sulfate proteoglycan, has been shown to promote CRC cell proliferation, migration, and invasion via epithelial-mesenchymal transition (EMT) and activation of the MAPK/ERK pathway58. This parallel underscore the broader relevance of the glycoprotein-MAPK signaling axis in colorectal tumorigenesis and supports the biological plausibility of our GSEA findings. Future studies employing advanced GSEA visualization tools, such as the GseaVis R package59, could further enhance the presentation and interpretation of pathway enrichment data. Our expanded analysis of the tumor immune microenvironment provides additional insights into the potential role of FJX1 in COAD. We observed that high FJX1 expression is associated with increased infiltration of M0 macrophages and neutrophils, coupled with decreased resting memory CD4+ T cells. Moreover, ssGSEA revealed positive correlations with immunosuppressive populations such as regulatory T cells and MDSCs60,61. These findings raise the intriguing possibility that FJX1 may not only drive intrinsic tumor cell aggressiveness via MAPK signaling but also modulate the extrinsic immune landscape to favor immune evasion62–64. Although the precise mechanisms remain to be elQ1ucidated, this dual role—if confirmed experimentally—could position FJX1 as an attractive target for combination therapies involving MAPK inhibitors and immune checkpoint blockade. This hypothesis warrants dedicated investigation in future studies. Several limitations must be acknowledged. First, the diagnostic 10-gene panel was not experimentally validated; only FJX1 was examined. Second, diagnostic performance metrics (e.g., ROC curves, sensitivity, specificity) were not calculated for the 10 hub genes; such analyses are essential for future validation studies. Third, our validation of FJX1 was limited to mRNA level by ISH; protein-level confirmation by immunohistochemistry was not feasible due to sample constraints. Given that FJX1 is a glycoprotein, future studies should prioritize IHC analysis in larger cohorts to assess FJX1 protein expression and, if possible, evaluate its glycosylation status. Fourth, the clinical cohort was relatively small (n = 39) and lacked survival follow-up data, precluding independent prognostic validation. Fifth, the mechanistic link between FJX1 and MAPK signaling is based solely on bioinformatic enrichment and requires experimental verification. Sixth, functional validation of FJX1’s oncogenic role in COAD cell lines (proliferation, migration, invasion assays) and in vivo models is lacking and represents a critical next step. Seventh, the PPI network was sparse, and hub gene selection may be influenced by the stringent interaction threshold; future studies using additional databases and experimental approaches could refine these findings. Eighth, single-cell RNA sequencing data could provide insights into the cellular localization of FJX1 expression, which we plan to explore in future work. Despite these limitations, our study provides a comprehensive bioinformatic analysis of glycoprotein-related genes in COAD and highlights FJX1 as a promising prognostic biomarker. Future research should focus on validating the 10-gene panel in large multicenter cohorts, performing protein-level and functional studies of FJX1, and exploring its potential as a therapeutic target. Specifically, in vitro experiments using FJX1 knockdown and overexpression in COAD cell lines (e.g., HCT116, SW480) are needed to assess effects on proliferation, migration, and invasion, as well as to confirm MAPK pathway activation via Western blotting for phosphorylated ERK, p38, and JNK. Transcriptomic profiling after FJX1 manipulation could further identify downstream effectors. Finally, investigating the interplay between FJX1, MAPK signaling, and immune evasion mechanisms may uncover novel therapeutic vulnerabilities in COAD. Conclusions In summary, we identified 10 glycoprotein-related hub genes as candidate diagnostic biomarkers warranting further validation and established FJX1 as an independent prognostic biomarker for COAD. FJX1 mRNA is overexpressed in tumors and correlates with advanced disease. Its association with MAPK signaling suggests a potential oncogenic role, and its correlation with immune infiltration patterns implies a possible contribution to the immunosuppressive tumor microenvironment. These findings provide a theoretical foundation for developing glycoprotein-based diagnostic panels and FJX1-informed prognostic models in COAD, though further validation is warranted. Supplementary Information Supplementary Information 1. Supplementary Information 2. Supplementary Information 3. Supplementary Information 4. Supplementary Information 5. Supplementary Information 6. Acknowledgements We thank the patients who participated in this study and the staff at the Department of Pathology, First Affiliated Hospital of Nanchang University, for technical assistance. Furthermore, we acknowledge the anonymous reviewers for their supportive and constructive contributions. Abbreviations Colon adenocarcinoma The cancer genome atlas database In situ hybridization Differentially expressed genes Carcinoembryonic antigen Carbohydrate antigen 19 − 9 Four-jointed box protein 1 Fold change Gene ontology Kyoto encyclopedia of genes and genomes Overall survival Tissue microarrays Formalin-fixed, paraffin-embedded Colorectal cancer Fecal immunochemical test Gene set enrichment analysis Extracellular matrix Cellular component Biological process Molecular function Transcription factor Genomic data commons Author contributions All of the authors have given their approval to the final version of the paper. HL: Writing - Original Draft, Methodology, Data curation, Experimentation. LW: Writing - Original Draft, Formal analysis, Validation. QF: Conceptualization, Supervision, Writing - Review & Editing. Funding The funding bodies did not engage in the study’s design, data management, or manuscript preparation. This research was supported by grants from the Science and Technology Program of the Administration of Traditional Chinese Medicine of Jiangxi Province (Grant Nos. 2023B1190 and 2024B0385) and the Science and Technology Program of the Health Commission of Jiangxi Province (Grant No. 202410118). Data availability The datasets utilized in this study are publicly accessible through the following repositories: The Cancer Genome Atlas Colon Adenocarcinoma (TCGA-COAD) data were sourced from the Genomic Data Commons (GDC) portal (https://portal.gdc.cancer.gov/). Additionally, a separate compilation of glycoprotein-related genes was derived from the UniProt database (https://www.uniprot.org/). It is important to note that this study did not incorporate the following types of data: protein sequences, DNA and RNA sequences, genetic polymorphisms, genotype-phenotype linkage data, macromolecular structures, microarray data (which must comply with MIAME standards), or crystallographic data for small molecules. Declarations Competing interests The authors declare no competing interests. Ethics approval This study was approved by the Ethics Committee of Jiangxi Provincial People’s Hospital, with the ethical approval number JXPPH-202310172. All procedures were conducted in strict accordance with the principles set forth in the Declaration of Helsinki. Consent to participate Informed consent was obtained from all individual participants involved in the study. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Information 1. Supplementary Information 2. Supplementary Information 3. Supplementary Information 4. Supplementary Information 5. Supplementary Information 6. Data Availability Statement The datasets utilized in this study are publicly accessible through the following repositories: The Cancer Genome Atlas Colon Adenocarcinoma (TCGA-COAD) data were sourced from the Genomic Data Commons (GDC) portal (https://portal.gdc.cancer.gov/). Additionally, a separate compilation of glycoprotein-related genes was derived from the UniProt database (https://www.uniprot.org/). It is important to note that this study did not incorporate the following types of data: protein sequences, DNA and RNA sequences, genetic polymorphisms, genotype-phenotype linkage data, macromolecular structures, microarray data (which must comply with MIAME standards), or crystallographic data for small molecules.