{"paper_id":"0c2c3bcd-d505-418a-b1d4-b1c8c0a0fd1c","body_text":"1\n1 Molecular and Structural Reprogramming of Gastric Cancer Revealed by Systems-Level \n2 Transcriptomic Analysis\n3\n4 Negar Mottaghi-Dastjerdi1*, Mohammad Soltany-Rezaee-Rad2\n5\n6 1 Department of Pharmacognosy and Pharmaceutical Biotechnology, School of Pharmacy, Iran \n7 University of Medical Sciences, Tehran, Iran\n8 2 Behestan Innovation Factory, Behestan Darou, Tehran, Iran\n9\n10 Corresponding Author: \n11 Dr. Negar Mottaghi-Dastjerdi\n12 Department of Pharmacognosy and Pharmaceutical Biotechnology, School of Pharmacy, Iran \n13 University of Medical Sciences, Tehran, Iran\n14 Tel (+98)21-44606181\n15 Fax: (+98)21-44606181\n16 E-mail: Mottaghi.n@iums.ac.ir\n17\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n2\n18 Abstract\n19 Gastric cancer (GC) is a major cause of cancer mortality and remains difficult to diagnose early \n20 and treat effectively. Although transcriptomic profiling has defined extensive molecular \n21 heterogeneity, many studies are not anchored to clinicopathological variables or interpreted in the \n22 context of tissue-level pathobiology. We applied an integrative transcriptomic and network-based \n23 framework to identify molecular signatures that reflect structural and biochemical reprogramming \n24 of gastric tissue during malignant transformation. RNA-seq expression profiles and clinical \n25 annotations were analyzed using non-parametric differential expression filtering, functional \n26 enrichment, and protein–protein interaction network modeling. Expression patterns were further \n27 evaluated across clinicopathological strata (stage, grade, nodal status, and metastasis) and by \n28 network-informed clustering. Across clinical strata, GC showed a consistent signature of \n29 developmental reactivation and loss of gastric epithelial identity. Developmental regulators, most \n30 prominently HOX-cluster genes and the histone variant HIST1H3J, were upregulated, consistent \n31 with epigenetic/chromatin reprogramming. In parallel, gastric differentiation and secretory lineage \n32 markers (ATP4A, KCNE2, PTF1A, VSTM2A) were persistently downregulated, reflecting \n33 suppression of parietal/ductal programs and altered metabolic/secretory function. ADIPOQ \n34 demonstrated stage-dependent repression and was associated with poorer survival, supporting a \n35 context-dependent prognostic role. Enrichment and network analyses also highlighted FGFR-\n36 centered signaling as a dominant oncogenic axis linked to proliferation and invasion. This study \n37 identifies a molecular pathology framework in which GC progression involves coordinated \n38 chromatin-level developmental reprogramming and sustained loss of gastric differentiation \n39 programs, accompanied by FGFR-driven oncogenic signaling and stage-dependent metabolic \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n3\n40 disruption. These signatures provide candidate diagnostic and prognostic biomarkers and support \n41 prioritization of therapeutically actionable pathways in GC.\n42 Keywords:  Biomarker discovery, Gastric cancer, Gene network analysis, HOX genes, Systems \n43 pathology, Transcriptomic profiling.\n44\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n4\n45 1. Introduction \n46 Gastric cancer (GC) remains as one of the main global health burden and ranks among the leading \n47 causes of cancer‑ related mortality, especially in East Asian populations where late‑stage diagnoses \n48 and therapeutic resistance are prevalent [1, 2]. Recent data indicate that GC remains highly \n49 prevalent across East Asia, with persistent high mortality and only modest improvements in \n50 survival over recent years [1, 3, 4].  Despite advances in surgical and chemotherapeutic \n51 intervention, the 5‑ year overall survival of GC remains below 30‑ 35% throughout much of the \n52 world, particularly with diagnoses at advanced stages [1, 5]. In the last decade, large‑scale \n53 transcriptomic profiling analysis, utilizing large datasets including those based on TCGA data and \n54 integrated bulk/single‑ cell datasets, has illuminated extensive molecular heterogeneity and \n55 complexity in GC, uncovering novel subtypes, signaling networks, and gene expression patterns \n56 [6, 7]. However, it is yet difficult to translate those findings into clinically valuable biomarkers \n57 and therapeutic targets.\n58 GC pathobiology involves coordinated transcriptional dysregulation, immune evasion, metabolic \n59 rewiring, and dedifferentiation. During tumorigenesis, gastric epithelial identity is progressively \n60 lost as embryonic/developmental pathways (e.g., Wnt, Notch, FGFR) are reactivated, producing \n61 transcriptomic signatures linked to early malignancy and clinical outcomes. Systems-biology and \n62 gene-network analyses help decode these complex patterns and prioritize central regulatory genes \n63 with translational relevance [8, 9].\n64 Despite extensive transcriptomic profiling, many studies still report DEGs without network \n65 context or stratification by key clinicopathological variables (stage, grade, nodal status). As a \n66 result, systems-level insight into developmental reactivation (e.g., HOX clusters) and loss of \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n5\n67 epithelial lineage programs remains limited. While integrative network modeling has proven \n68 useful for identifying drivers and resistance mechanisms across GI and hepatic cancers [10-16], \n69 unified pipelines that link phenotype correlations with multidimensional network clustering are \n70 still uncommon, leaving a gap at the intersection of transcriptomics, network centrality, and \n71 clinical relevance for biomarker discovery.\n72 To address these gaps, we performed an integrated transcriptomic and systems-level analysis of \n73 TCGA-STAD RNA-seq data, combining non-parametric DEG filtering with dual PPI network \n74 construction, functional enrichment, and clinicopathological stratification (stage, grade, nodal \n75 status). This framework aimed to identify stable diagnostic and prognostic biomarkers, \n76 characterize developmental and immune transcriptional programs, and prioritize therapeutic \n77 candidates, particularly within HOX genes, gastric lineage markers, and core regulatory \n78 transcription factors—building on our prior work in gastric and other gastrointestinal cancers [17, \n79 18].\n80 2. Materials and methods\n81 2.1. Computational environment and tools\n82 All preprocessings of RNA-seq data, normalization, filtering, and differential expression analysis \n83 were carried out in Python (v3.10) via Jupyter Notebook. Main involved libraries were pandas for \n84 data manipulation, numpy for mathematical calculations, and matplotlib and seaborn for data \n85 plotting purposes.\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n6\n86 2.2. Filtering low-expression genes and log2 normalization\n87 Raw counts of 448 TCGA gastric cancer samples were removed for low-expression genes (≤1 \n88 count in >95% of the samples), and 56,678 genes expressed in ≥5 samples were left. Log₂(x + 1) \n89 transformation was then applied for stabilization of variance and for enhancing comparability.\n90 2.3. Clinical metadata processing \n91 Clinical data for 511 TCGA-STAD samples were retrieved from GDC, and key variables (e.g., ID, \n92 gender, stage, survival) were extracted. Data were cleaned and standardized using Python (v3.10) \n93 and pandas. \n94 2.4. Differential expression analysis (DEA)\n95 RNA-seq data from 448 samples (412 tumors and 36 normal tissues) were analyzed using the \n96 Mann–Whitney U test. Differentially expressed genes (DEGs) were defined as those with |log2FC| \n97 ≥ 1 and FDR < 0.05 (Benjamini–Hochberg). Genes with an average expression ≥ 1 in both groups \n98 were retained, and the top 100 upregulated and top 100 downregulated DEGs were selected for \n99 downstream analyses. Ensembl IDs were converted to gene symbols using MyGene.info. All \n100 analyses were performed in Python (v3.10) using the pandas, scipy, statsmodels, and mygene \n101 packages. The complete DEG list is provided in Supplementary File 1.\n102 2.5. Functional enrichment analysis\n103 Upregulated (n=85) and downregulated (n=94) DEGs were analyzed using ENRICHR to identify \n104 enriched Gene Ontology (GO: Biological Process (BP), Molecular Function (MF), and Cellular \n105 Component (CC)) terms and Reactome 2024 pathways. Only results with FDR < 0.05 were \n106 considered significant. Enrichment results were visualized as bar plots based on −log₁₀(FDR) for \n107 comparative interpretation.  \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n7\n108 2.6. Protein–protein interaction (PPI) network analysis\n109 Two PPI networks were constructed using STRING in Cytoscape v3.10: one from ~6,500 DEGs \n110 (split into four subsets) and another from the top 100 up- and downregulated genes. Networks were \n111 generated with a confidence cutoff of 0.15. Hub genes were identified using cytoHubba via MCC, \n112 DMNC, MNC, and Degree methods; genes ranked top-5 across multiple metrics were deemed \n113 high-confidence hubs.\n114 2.7. Cluster analysis of the network\n115 To identify functional modules within the 6,500-gene PPI network, we applied the IPCA algorithm \n116 in CytoCluster (Cytoscape v2.1.0) [19] with TinThreshold = 0.9, ComplexSize ≥ 10, and \n117 PathLength = 2. This yielded 28 clusters. The top 4 clusters (ranked by density) were analyzed for \n118 pathway enrichment using Reactome via Enrichr, focusing on the top 20 significant pathways per \n119 cluster.\n120 2.8. UALCAN-based clinicopathological expression and survival analysis\n121 Expression of the 13 hub genes was assessed in TCGA-STAD using the UALCAN portal \n122 (https://ualcan.path.uab.edu). TPM values were retrieved for cancer stage, tumor grade, nodal \n123 metastasis status, and survival modules, with normal gastric tissue as the reference. UALCAN \n124 applies Welch’s unpaired t-tests for subgroup comparisons, and Kaplan–Meier curves were used \n125 to evaluate prognostic associations (P < 0.05).\n126 3. Results \n127 3.1. Identification of DEGs in GC\n128 RNA-seq data of 448 gastric samples (412 tumor, 36 normal) was compared with the Mann–\n129 Whitney U test to identify the DEGs. The genes with a threshold of |log2FC| ≥ 1 and FDR < 0.05 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n8\n130 were designated as significant, and there were 9,896 DEGs (9,592 upregulated, 304 \n131 downregulated). To improve reliability, we retained only genes with average expression ≥ 1 in \n132 both groups. From the filtered list, the top 100 upregulated and top 100 downregulated DEGs (by \n133 absolute log2FC) were taken for downstream analysis. Gene IDs were converted to official \n134 symbols using the MyGene.info API. \n135 3.2. Network reconstruction and analysis\n136 To distinguish therapeutic targets from diagnostic/prognostic biomarkers, we constructed two PPI \n137 networks from TCGA-STAD DEGs. The genome-scale network (~6,500 DEGs) mapped global \n138 GC interactions and, using cytoHubba (four topology algorithms), identified highly central hubs, \n139 including histone variants (H3C12, H3-4), transcription factors (PAX2, HOXD11/13, HOXC12), \n140 signaling molecules (PRKACG, FGF8), and immune mediators (IFNG, IFNL2), as candidate drug \n141 targets because perturbing them could disrupt broad tumor-promoting circuitry (Table 1). \n142 Table 1. Hub genes from the PPI network of ~6,500 DEGs, prioritized based on overlap across \n143 four cytoHubba topological methods. Gene locations and aliases are provided for annotation. \n# Node LogFC Gene \nID\nGene Description Rank in \nEach \nRanking \nMethod\nOther names Location \n1 H3C12 1.6 8356 H3 clustered \nhistone 12\nMCC (1), \nMNC (3), \nDegree (3)\nH3/j; H3C1; H3C2; \nH3C3; H3C4; H3C6; \nH3C7; H3C8; H3FJ; \nH3C10; H3C11; \nHIST1H3J\n6p22.1\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n9\n2 H3-4 1.19 8290 H3.4 histone, \ncluster member\nMCC (2), \nMNC (2), \nDegree (2)\nH3t; H3.4; H3/g; \nH3FT; H3C16; \nHIST3H3\n1q42.13\n3 PAX2 1.3 5076 paired box 2 MCC (3) FSGS7; PAPRS; \nPAX-2\n10q24.31\n4 HOXD11 1.6 3237 homeobox D11 MCC (4), \nDMNC (1)\nHOX4; HOX4F 2q31.1\n5 HOXD13 1.4 3239 homeobox D13 MCC (5) BDE; SPD; BDSD; \nSPD1; HOX4I\n2q31.1\n6 PRKACG 1.6 5568 protein kinase \ncAMP-activated \ncatalytic subunit \ngamma\nMNC (1), \nDegree (1)\nKAPG; PKACg; \nBDPLT19\n9q21.11\n7 FGF8 1.02 2253 fibroblast growth \nfactor 8\nMNC (4), \nDegree (4)\nHH6; AIGF; KAL6; \nFGF-8; HBGF-8\n10q24.32\n8 IFNG 1.03 3458 interferon gamma MNC (5), \nDegree (5)\nIFG; IFI; IMD69 12q15\n9 OR8J1 3.7 219477 olfactory receptor \nfamily 8 \nsubfamily J \nmember 1\nDMNC (2) OR11-183 11q12.1\n10 OR5B21 2.22 219968 olfactory receptor \nfamily 5 \nsubfamily B \nmember 21\nDMNC (3) - 11q12.1\n11 HOXC12 3.06 3228 homeobox C12 DMNC (4) HOX3; HOC3F; \nHOX3F\n12q13.13\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n10\n12 IFNL2 3.02 282616 interferon lambda \n2\nDMNC (5) IL28A; IFNL2a; \nIFNL3a; IL-28A\n19q13.2\n144 A second, focused network built from the top 200 DEGs (100 up, 100 down) yielded 93 nodes and \n145 318 edges (Fig 1) and highlighted 13 hubs (HOXC8, HOXC9, HOXA11, HOXC11, HOXA13, \n146 WT1, ADIPOQ, GCG, VSTM2A, ATP4A, KCNE2, PTF1A) with both strong expression shifts \n147 and network centrality, supporting their potential as diagnostic/prognostic biomarkers (Table 2). \n148 This dual-network strategy separates broadly actionable targets from clinically tractable \n149 biomarkers by integrating systems-level topology with differential expression magnitude.\n150 Table 2. Hub genes from the PPI network of the top 200 DEGs, including ranking methods and \n151 log2 fold change (logFC) values.\n# Node LogFC Gene \nID\nGene Description Rank Ranking \nMethod\nOther names Location \n1 H3C12 1.6 8356 H3 clustered \nhistone 12\n1, 1, \n1\nMCC, \nMNC, \nDegree\nH3/j; H3C1; \nH3C2; H3C3; \nH3C4; H3C6; \nH3C7; H3C8; \nH3FJ; H3C10; \nH3C11; HIST1H3J\n6p22.1\n2 HOXC8 1.6 3224 homeobox C8 2 MCC HOX3; HOX3A 12q13.13\n3 HOXC9 2.1 3225 homeobox C9 3 MCC HOX3; HOX3B 12q13.13\n4 HOXA11 2.2 3207 homeobox A11 4 MCC HOX1; HOX1I; \nRUSAT1\n7p15.2\n5 HOXC11 1.7 3227 homeobox C11 5, 1 MCC, \nDMNC\nHOX3H 12q13.13\n6 HOXA13 1.5 3209 homeobox A13 2 DMNC HOX1; HOX1J 7p15.2\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n11\n7 PTF1A -1.2 256297 pancreas associated \ntranscription factor \n1a\n3 DMNC p48; PACA; \nPAGEN2; \nbHLHa29; PTF1-\np48\n10p12.2\n8 VSTM2A -1.2 222008 V-set and \ntransmembrane \ndomain containing \n2A\n4 DMNC VSTM2 7p11.2\n9 KCNE2 -1.2 9992 potassium voltage-\ngated channel \nsubfamily E \nregulatory subunit \n2\n5 DMNC LQT5; LQT6; \nATFB4; MIRP1\n21q22.11\n10 GCG -1.2 2641 glucagon 2, 2 MNC, \nDegree\nGLP1; GLP2; \nGRPP; GLP-1\n2q24.2\n11 WT1 1.4 7490 WT1 transcription \nfactor\n3, 4 MNC, \nDegree\nGUD; AWT1; \nWAGR; WT-1; \nWT33; NPHS4; \nWIT-2\n11p13\n12 ADIPOQ -1.2 9370 adiponectin, C1Q \nand collagen \ndomain containing\n3, 5 MNC, \nDegree\nACDC; ADPN; \nAPM1; APM-1; \nGBP28; ACRP30; \nADIPQTL1\n3q27.3\n13 ATP4A -1.2 495 ATPase H+/K+ \ntransporting subunit \nalpha\n3, 3 MNC, \nDegree\nATP6A 19q13.12\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n12\n152 3.3. Functional enrichment analysis\n153 Functional enrichment analysis was performed to understand the biological role of identified up- \n154 and downregulated genes. GO terms and pathway associations were analyzed in two sets \n155 separately to disclose different functional profiles of both gene sets (Fig 2 and 3).\n156 Upregulated genes (n = 85) were enriched for embryonic/developmental programs and immune \n157 differentiation (e.g., morphogenesis terms and myeloid dendritic cell activation), along with \n158 transcriptional regulation (sequence-specific DNA binding) and multiple solute/anion transport \n159 activities. Reactome analysis similarly highlighted organic anion and vitamin/nucleoside transport, \n160 kidney developmental pathways, and drug-metabolism modules (e.g., ciprofloxacin/atorvastatin \n161 ADME, bile acid metabolism), supporting a phenotype of developmental reactivation, \n162 transcriptional control, and altered transport/metabolic capacity.\n163 Downregulated genes were predominantly enriched for ion homeostasis and membrane transport, \n164 particularly sodium/potassium handling (transmembrane transport, export/import, membrane \n165 repolarization), and localized to vesicular/endosomal–Golgi compartments and transport \n166 complexes (e.g., clathrin-coated vesicles, Na⁺/K⁺ -ATPase complexes). Functional terms also \n167 indicated reduced channel and transporter activities, with Reactome showing decreased innate \n168 immune defense (defensins/antimicrobial peptides), aquaporin and ion channel transport, and \n169 neuroendocrine signaling pathways (acetylcholine release, incretin/glucagon signaling), consistent \n170 with loss of normal physiological and immune functions.\n171 Overall, GC displayed a shift toward developmental reprogramming, transcriptional activation, \n172 and altered solute/drug metabolism (upregulated genes), alongside suppression of epithelial \n173 ion/transport homeostasis and innate immunity (downregulated genes), underscoring tissue \n174 remodeling and functional derangement relevant to biomarker and target discovery.\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n13\n175 3.4. Cluster analysis and reactome pathway enrichment\n176 We further enriched the top four clusters obtained from IPCA into Reactome pathways to gain \n177 deeper insights into the modular organization of the gene network and to identify distinct \n178 biological themes. These clusters were mainly enriched for FGFR signaling and its canonical \n179 downstream pathways, developmental biology, and transcriptional regulation, showing that they \n180 are biologically coherent and functionally specialized (Fig 4, and Table 3).\n181 Table 3. Classification of Reactome Pathways Enriched Across Top Four Clusters.\nCategory Reactome Pathway Clusters\nFGFR Signaling FGFR1 Ligand Binding and Activation 1, 2, 3\nFGFR1c Ligand Binding and Activation 4\nFGFR2 Ligand Binding and Activation 1, 2, 3\nFGFR2c Ligand Binding and Activation 4\nFGFR3 Ligand Binding and Activation 4\nFGFR3b Ligand Binding and Activation 4\nFGFR3c Ligand Binding and Activation 4\nFGFR4 Ligand Binding and Activation 4\nFGFR2 Mutant Receptor Activation 1, 2, 3\nFGFR3 Mutant Receptor Activation 4\nActivated Point Mutants of FGFR2 1, 2, 3\nActivated Point Mutants of FGFR1 4\nActivated Point Mutants of FGFR3 4\nSignaling by FGFR2 in Disease 2, 3\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n14\nFGFRL1 Modulation of FGFR1 Signaling 4\nNegative Regulation of FGFR1 Signaling 1, 2, 3\nNegative Regulation of FGFR2 Signaling 1, 2, 3\nCanonical Signaling \nCascades\nSHC-mediated Cascade FGFR1 1, 2, 3\nSHC-mediated Cascade FGFR2 1, 2, 3\nFRS-mediated FGFR1 Signaling 1, 2, 3\nFRS-mediated FGFR2 Signaling 1, 2, 3\nPhospholipase C-mediated Cascade FGFR1 1, 2, 3\nPhospholipase C-mediated Cascade; FGFR2 1, 2, 3\nPhospholipase C-mediated Cascade; FGFR3 4\nPI-3K Cascade FGFR1 1, 2, 3\nPI-3K Cascade FGFR2 1, 2, 3\nPI3K Cascade 2, 3\nIRS-mediated Signalling 3\nDownstream Signaling of Activated FGFR1 1, 2, 3\nDownstream Signaling of Activated FGFR2 1, 2, 3\nDevelopmental and \nOrganogenesis\nDevelopmental Biology 1, 2, 4\nGastrulation 1, 2, 4\nKidney Development 1\nNephron Development 4\nFormation of the Ureteric Bud 1\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n15\nFormation of Intermediate Mesoderm 4\nFormation of the Anterior Neural Plate 4\nFormation of the Posterior Neural Plate 4\nSpecification of the Neural Plate Border 4\nTranscriptional Regulation Regulation of Gene Expression in Early Pancreatic \nPrecursor Cells\n4\nRegulation of Expression of SLITs and ROBOs 4\nMatrix Remodeling and \nInvasion\nActivation of Matrix Metalloproteinases 3\n182\n183 Across all clusters, FGFR signaling emerged as a consistent hallmark, including ligand-driven \n184 activation of FGFR1–4 and isoforms (FGFR1c/2c/3b/3c), alongside evidence of oncogenic \n185 FGFR2 mutant activation and regulatory counterbalance via negative modulators (e.g., FGFRL1). \n186 Enrichment extended to major downstream cascades, SHC/FRS-mediated signaling, PLC, and \n187 PI3K/IRS pathways, implicating proliferation, survival, motility, and metabolic regulation. \n188 Developmental programs were also reactivated (e.g., gastrulation, kidney/nephron and ureteric bud \n189 development, neural plate formation), suggesting increased cellular plasticity and possible EMT-\n190 related microenvironmental shifts. Additional cluster-specific themes included disrupted \n191 transcriptional control (early pancreatic precursor programs; SLIT/ROBO regulation) and matrix \n192 remodeling via metalloproteinase activation. Overall, the IPCA clusters define a coherent network \n193 in which FGFR signaling links developmental reprogramming, intracellular signaling, \n194 transcriptional regulation, and invasive behavior in gastric cancer. \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n16\n195 3.5. Analysis of tumor grade-specific gene expression and potential biomarker utility\n196 Gene expression analysis across tumor grades in stomach adenocarcinoma (STAD) uncovered a \n197 dynamic transcriptional landscape, with several genes exhibiting significant and grade-specific \n198 expression changes (Fig 5). These alterations offer insight into molecular mechanisms underlying \n199 tumor progression and differentiation, and highlight potential diagnostic and prognostic \n200 biomarkers.\n201 Grade-stratified analysis showed increasing expression of \n202 HOXA11/HOXA13/HOXC8/HOXC9/HOXC11 and HIST1H3J from Grade 1 to Grades 2–3, \n203 consistent with developmental/epigenetic reprogramming. In contrast, gastric lineage markers \n204 (ATP4A, KCNE2, PTF1A, VSTM2A) were strongly downregulated early and remained low \n205 across grades, indicating sustained loss of epithelial identity. WT1 rose mainly in Grades 2–3, \n206 while ADIPOQ was repressed early with modest recovery and GCG declined gradually. Together, \n207 these patterns support a panel where HOX/HIST1H3J (±WT1) reflect progression, and \n208 ATP4A/KCNE2/PTF1A/VSTM2A (±ADIPOQ) mark early dedifferentiation.\n209 3.6. Expression of genes in STAD based on nodal metastasis status and biomarker potential\n210 When TCGA-STAD samples were examined according to nodal metastasis stage (N0–N3), a clear \n211 and recurring alteration appeared among the 13 studied genes, pointing to their possible value as \n212 biomarkers for identifying and tracking disease (Fig 6).\n213 Early Detection and Diagnostic Biomarkers: The HOX cluster genes (HOXA11, HOXA13, \n214 HOXC8, HOXC9, and HOXC11) together with HIST1H3J, showed marked and statistically \n215 strong overexpression across every nodal category when compared with normal gastric tissue \n216 (most with P ≤ 10⁻ ⁹). The elevated expression was already present in node-negative tumors, \n217 pointing to early engagement of these transcription factors in tumor development rather than \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n17\n218 secondary activation during metastasis. Seen in this light, their persistent activity may serve as an \n219 early warning sign for primary tumors that have yet to reach the lymphatic system.\n220 Progression and Metastatic Biomarkers:  While the expression of HOX genes and HIST1H3J did \n221 not show a progressive increase with higher nodal stages (N1–N3), certain genes demonstrated \n222 patterns suggestive of progression markers. Expression of WT1 rose sharply in N0 and N1 tumors \n223 (P around 10⁻⁶–10⁻⁷) and climbed even higher by the N3 stage, which may point to its role in \n224 promoting or tracking advanced nodal spread. The smaller but steady changes seen for KCNE2 \n225 and PTF1A across the same stages might reflect the gradual erosion of differentiation as tumors \n226 become more metastatic, an observation that deserves closer study for its potential prognostic \n227 value.\n228 Loss of Differentiation Markers: Gastric lineage genes (ATP4A, KCNE2, PTF1A, VSTM2A) and \n229 ADIPOQ were markedly downregulated in both node-negative and metastatic tumors (all P < 0.01 \n230 vs normal), indicating early, sustained dedifferentiation and supporting their use as negative \n231 diagnostic markers independent of nodal status. GCG showed no nodal-group differences and can \n232 be excluded from the core biomarker set. Overall, the data support a two-tier framework: early \n233 activation markers (HOX-cluster genes, HIST1H3J), persistent loss-of-identity markers (ATP4A, \n234 KCNE2, PTF1A, VSTM2A, ADIPOQ), and WT1 as a potential progression marker associated \n235 with nodal spread.\n236 3.7. Tumor stage–specific gene expression and implications for biomarker discovery in STAD\n237 As we looked across tumor stages in STAD, certain genes stood out for how sharply their \n238 expression shifted; patterns that could eventually guide biomarker discovery and treatment design. \n239 Several genes, among them HOXA11, HOXA13, HOXC8, HOXC9, HOXC11, HIST1H3J, WT1, \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n18\n240 and the differentiation markers ATP4A, KCNE2, VSTM2A, and PTF1A, showed stage-specific \n241 shifts in activity, suggesting value for classifying tumors or refining clinical decisions (Fig 7).\n242 Across STAD stages, HIST1H3J and multiple HOX-cluster genes are consistently upregulated \n243 (often P < 10⁻¹²), with early onset and sustained expression, suggesting roles in tumor \n244 establishment and maintenance and potential utility as early molecular markers. In contrast, gastric \n245 differentiation genes ATP4A, KCNE2, VSTM2A, and PTF1A are persistently downregulated \n246 across all stages versus normal tissue, indicating stable loss of parietal/ductal identity and \n247 supporting their use as diagnostic markers of malignant transformation. WT1 increases with stage, \n248 reaching stronger significance in advanced disease, consistent with a progression-associated \n249 biomarker. Together, these patterns define a stable STAD signature, early HOX/HIST1H3J \n250 activation with sustained repression of gastric identity genes, while WT1 may aid risk stratification \n251 and monitoring.\n252 3.8. Survival analysis of candidate genes in STAD\n253 We used Kaplan–Meier analysis to examine how the 13 candidate genes relate to patient survival \n254 in STAD. Of all the genes tested, only ADIPOQ showed a significant link with overall survival (p \n255 = 0.012); the others showed no clear association (p > 0.05) (Fig 8).\n256 Patients were stratified by ADIPOQ expression (high, n = 100 vs low/intermediate, n = 292). \n257 Kaplan–Meier analysis showed poorer survival in the high-expression group, suggesting ADIPOQ \n258 as a negative prognostic marker in STAD. As adiponectin modulates metabolism and \n259 inflammation, this pattern supports a context-dependent role in the tumor microenvironment. \n260 Clinically, ADIPOQ may aid risk stratification, but independent cohort validation and mechanistic \n261 studies are needed.\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n19\n262 4. Discussion\n263 In this work, we took a broad look at the transcriptome of STAD to map gene expression changes \n264 and their biological and clinical implications. By combining a full set of roughly 6,500 DEGs with \n265 a focused analysis of the top 200, two clear interaction modules began to emerge from the data. \n266 The first includes histone genes, immune regulators, and kinases that occupy central positions in \n267 the protein–protein interaction network, making them attractive targets for therapeutic disruption. \n268 The second group includes transcription factors, epithelial genes, and metabolic regulators that \n269 show striking expression differences. These features make them appealing candidates for \n270 developing both diagnostic and prognostic markers.\n271 We found that HOXA11, HOXA13, HOXC8, HOXC9, and HOXC11 were consistently more \n272 active across tumor grades, stages, and nodal groups, a result that fits well with earlier evidence of \n273 HOX gene reactivation in gastrointestinal cancers [20]. Their expression tended to climb as tumors \n274 became less differentiated and more advanced, most noticeably in the aggressive forms. This \n275 steady rise links them to both loss of cellular identity and increasing malignancy. The histone gene \n276 HIST1H3J showed a comparable pattern, its levels rose step by step with tumor grade and stage, \n277 which aligns with reports of broad epigenetic remodeling [21]. These genes seem to switch on \n278 early in tumor formation and remain active as the disease develops, helping to sustain \n279 transcriptional activity that supports growth and survival.\n280 Conversely, the observed downregulation of gastric lineage markers (ATP4A, KCNE2, VSTM2A, \n281 PTF1A) from early stages supports the idea that dedifferentiation is an initiating event in gastric \n282 tumorigenesis [22, 23]. Our observations agree with earlier reports showing that genes tied to acid \n283 secretion and epithelial differentiation are already dampened in the early stages of gastric cancer. \n284 In particular, several studies have noted reduced expression of ATP4A and ATP4B, both of which \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n20\n285 have been examined as potential diagnostic markers [22]. ADIPOQ also showed dynamic \n286 regulation, with an early decrease followed by partial reactivation in advanced tumors; recent work \n287 suggests that low adiponectin levels or expression correlate with more aggressive GC features \n288 [24]. Survival analysis revealed that high ADIPOQ expression correlates with poorer patient \n289 outcomes, further highlighting its potential as a context-dependent prognostic biomarker [24].\n290 Our enrichment analysis pointed to a set of molecular patterns that mirror what other \n291 transcriptomic studies have described in gastric cancer. The genes showing higher expression were \n292 mostly involved in transcriptional control, immune signaling, and developmental processes, \n293 similar to earlier reports describing a reactivation of embryonic and inflammatory pathways during \n294 tumor growth [25]. By contrast, the genes showing reduced activity were mostly linked to ion \n295 transport, vesicle movement, and antimicrobial defense. Their decline mirrors the loss of normal \n296 gastric cell function that tends to accompany malignant transformation [26]. Overall, the data \n297 suggest a broad reorganization of the tumor transcriptome, shifting away from the physiological \n298 roles of the stomach toward a more adaptable, development-like and metabolically flexible state, \n299 a trend that recent single-cell and spatial transcriptomic analyses of gastric and gastrointestinal \n300 cancers have also captured [25, 26]. Our cluster-based pathway enrichment further emphasized the \n301 centrality of FGFR signaling and its downstream cascades. This finding is in agreement with recent \n302 clinical and preclinical studies demonstrating FGFR dysregulation as a driver of oncogenic \n303 transcriptional programs in gastric cancer [27, 28].\n304 A key strength of this study is its layered design, combining stringent DEG filtering with network \n305 modeling, functional annotation, and clinicopathological interpretation to identify clinically \n306 relevant targets. Our findings are consistent with recent bulk and single-cell GC studies reporting \n307 metabolic reprogramming, immune microenvironment changes, and loss of gastric differentiation \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n21\n308 [7, 29]. Limitations include reliance on TCGA-STAD only, lack of experimental validation, and \n309 the inability of transcriptomics to capture post-transcriptional/post-translational regulation; future \n310 work should validate results in independent cohorts and integrate additional omics layers.\n311 Going forward, it will be important to check how well these biomarkers perform in other patient \n312 cohorts and to see whether they can be tracked in easier-to-obtain materials like blood or gastric \n313 fluid. Testing how HOX genes, WT1, and FGFR-pathway effectors behave when perturbed in cells \n314 or organoids should also help clarify what roles they actually play in tumor biology. Additionally, \n315 the relationship between ADIPOQ expression and the immune microenvironment warrants \n316 investigation to decipher its dual role in metabolism and immunoregulation.\n317 In summary, our study delineates two functionally distinct gene modules in gastric cancer: a \n318 regulatory core enriched in HOX and histone genes that may be exploited for early diagnosis and \n319 a set of suppressed differentiation genes marking the loss of gastric identity. These results outline \n320 a molecular framework that could guide biomarker-based patient classification and help shape new \n321 therapeutic strategies in STAD. They also highlight how combining network-level analysis with \n322 clinical data can move the field closer to truly precise cancer care.\n323 5. Conclusion \n324 Using an integrated transcriptomic and network-based framework (differential expression, PPI \n325 mapping, hub ranking, and pathway enrichment), we identified a coordinated STAD signature \n326 marked by upregulation of developmental/immune programs and repression of epithelial \n327 differentiation and transport genes. Dual-network analysis prioritized a focused biomarker set—\n328 HOX genes, HIST1H3J, ATP4A, KCNE2, and PTF1A—with potential utility for early detection, \n329 monitoring, and risk stratification, and highlighted actionable pathways relevant to precision \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n22\n330 oncology. Key limitations are reliance on TCGA-STAD data and lack of experimental validation; \n331 future work should validate these markers in independent cohorts and test mechanisms in vitro/in \n332 vivo, ideally integrating proteomic/epigenomic and spatial transcriptomic layers to refine \n333 heterogeneity and clinical translation.\n334 Abbreviations\n335 BP: Biological Process; CC: Cellular Component; DEG: Differentially Expressed Gene; FGFR: \n336 Fibroblast Growth Factor Receptor; GC: Gastric Cancer; GO: Gene Ontology; HOX: Homeobox \n337 Gene; KEGG: Kyoto Encyclopedia of Genes and Genomes; MF: Molecular Function; PPI: \n338 Protein–Protein Interaction; RNA-seq: RNA Sequencing; STAD: Stomach Adenocarcinoma; \n339 TCGA: The Cancer Genome Atlas; UALCAN: University of Alabama at Birmingham Cancer \n340 Data Analysis Portal.\n341 Declarations\n342 Funding\n343 The authors received no specific funding for this work.\n344 Ethics statement\n345 This study used only publicly available, de-identified data and did not require ethics approval. \n346 Consent for publication\n347 Not applicable.\n348 Data Availability\n349 TCGA-STAD RNA-seq and clinical data are publicly available from the Genomic Data Commons \n350 (GDC). All processed data supporting the findings of this study are available at GitHub \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n23\n351 (https://github.com/negmot/tcga-stad-network-analysis) and have been archived on Zenodo (DOI: \n352 10.5281/zenodo.18537980). \n353 Competing interests\n354 The authors have no competing interests to declare that are relevant to the content of this article.\n355 Author contributions\n356 N.M.D.: Conceptualization, Supervision, Data curation, Formal analysis, Visualization, Writing-\n357 original draft.\n358 M.S.R.R.: Data curation, Investigation, Writing-review & editing.\n359 Both authors: Final approval of the manuscript.\n360 Supporting information\n361 Supplementary File 1. Full differential expression results tables (RAR/XLSX).\n362 References \n363 1. Chen Y, Jia K, Xie Y, Yuan J, Liu D, Jiang L, et al. The current landscape of gastric \n364 cancer and gastroesophageal junction cancer diagnosis and treatment in China: a comprehensive \n365 nationwide cohort analysis. J Hematol Oncol. 2025;18(1):42. doi: 10.1186/s13045-025-01698-y.\n366 2. Mousavi SE, Ilaghi M, Elahi Vahed I, Nejadghaderi SA. Epidemiology and \n367 socioeconomic correlates of gastric cancer in Asia: results from the GLOBOCAN 2020 data and \n368 projections from 2020 to 2040. Sci Rep. 2025;15(1):6529. doi: 10.1038/s41598-025-90064-6.\n369 3. Lin J-L, Lin J-X, Lin G-T, Huang C-M, Zheng C-H, Xie J-W, et al. Global incidence and \n370 mortality trends of gastric cancer and predicted mortality of gastric cancer by 2035. BMC Public \n371 Health. 2024;24(1):1763. doi: 10.1186/s12889-024-19104-6.\n372 4. Shin WS, Xie F, Chen B, Yu P, Yu J, To KF, et al. Updated Epidemiology of Gastric \n373 Cancer in Asia: Decreased Incidence but Still a Big Challenge. Cancers (Basel). \n374 2023;15(9):2639. doi: 10.3390/cancers15092639.\n375 5. Zhang Z, Wang J, Song N, Shi L, Du J. The global, regional, and national burden of \n376 stomach cancer among adolescents and young adults in 204 countries and territories, 1990–2019: \n377 A population-based study. Front Public Health. 2023;11:1079248. doi: \n378 10.3389/fpubh.2023.1079248.\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n24\n379 6. Zhang H, Yang W, Tan X, He W, Zhao L, Liu H, et al. Long-term relative survival of \n380 patients with gastric cancer from a large-scale cohort: a period-analysis. BMC Cancer. \n381 2024;24(1):1420. doi: 10.1186/s12885-024-13141-5.\n382 7. Lin X, Yang P, Wang M, Huang X, Wang B, Chen C, et al. Dissecting gastric cancer \n383 heterogeneity and exploring therapeutic strategies using bulk and single-cell transcriptomic \n384 analysis and experimental validation of tumor microenvironment and metabolic interplay. Front \n385 Pharmacol. 2024;15:1355269. doi: 10.3389/fphar.2024.1355269.\n386 8. Mottaghi-Dastjerdi N, Ghorbani A, Montazeri H, Guzzi PH. A systems biology approach \n387 to pathogenesis of gastric cancer: gene network modeling and pathway analysis. BMC \n388 Gastroenterol. 2023;23(1):248. doi: 10.1186/s12876-023-02891-4.\n389 9. Khoshdel F, Mottaghi-Dastjerdi N, Yazdani F, Salehi S, Ghorbani A, Montazeri H, et al. \n390 CTGF, FN1, IL-6, THBS1, and WISP1 genes and PI3K-Akt signaling pathway as prognostic and \n391 therapeutic targets in gastric cancer identified by gene network modeling. Discov Oncol. \n392 2024;15(1):344. doi: 10.1007/s12672-024-01225-4.\n393 10. Rahimi-Farsi N, Shahbazi T, Ghorbani A, Mottaghi-Dastjerdi N, Yazdani F, Mohseni P, \n394 et al. Network-based analysis of candidate oncogenes and pathways in hepatocellular carcinoma. \n395 Biochem Biophys Rep. 2025;43:102086. doi: 10.1016/j.bbrep.2025.102086.\n396 11. Alvani A, Mottaghi-Dastjerdi N, Gholami A, Ghorbani A, Pazhoohesh Z, Pajdam M, et \n397 al. Identification of indirect pathways enhancing the biocompatibility of DOX/GO/Fe3O4 \n398 nanomaterials in Glioblastoma: Gene network modeling and pathway analysis. Biochem Biophys \n399 Res Commun. 2025;773:152077. doi: 10.1016/j.bbrc.2025.152077.\n400 12. Rahimi-Farsi N, Ghorbani A, Mottaghi-Dastjerdi N, Shahbazi T, Bostanian F, Mohseni \n401 P, et al. Comprehensive system biology analysis of microRNA-101-3p regulatory network \n402 identifies crucial genes and pathways in hepatocellular carcinoma. J Genet Eng Biotechnol. \n403 2025;23(1):100471. doi: 10.1016/j.jgeb.2025.100471.\n404 13. Rahimi H, Farzadifar E, Mottaghi-Dastjerdi N, Ghorbani A, Amerizadeh F, Pasdar A. \n405 Integrative Systems Biology Analysis of MicroRNA Regulation in Wnt Signaling Pathway \n406 During Breast Cancer Progression. Int J Cancer Manag. 2025;18(18):e164192. doi: \n407 10.5812/ijcm-160657.\n408 14. Saeidi M, Pasdar A, Rahmani F, Ghorbani A, Mottaghi N, Amerizadeh F. A Prospective \n409 to Regulatory Role of miRNAs on Wnt/β-catenin Signaling and Its Crosstalk to the Other \n410 Cellular Pathways in Tumorigenesis of Glioblastoma by a Systems Biology Approach. Int J \n411 Cancer Manag. 2025;18(1):e156834. doi: 10.5812/ijcm-156834.\n412 15. Yazdani F, Mottaghi-Dastjerdi N, Shahbazi B, Ahmadi K, Ghorbani A, Soltany-Rezaee-\n413 Rad M, et al. Identification of key genes and pathways involved in T-DM1-resistance in OE-19 \n414 esophageal cancer cells through bioinformatics analysis. Heliyon. 2024;10(18):e37451. doi: \n415 10.1016/j.heliyon.2024.e37451.\n416 16. Soltany-Rezaee-Rad M, Mottaghi-Dastjerdi N, Setayesh N, Roshandel G, Ebrahimifard \n417 F, Sepehrizadeh Z. Overexpression of FOXO3, MYD88, and GAPDH identified by suppression \n418 subtractive hybridization in esophageal cancer is associated with autophagy. Gastroenterol Res \n419 Pract. 2014;2014(1):185035. doi: 10.1155/2014/185035.\n420 17. Mottaghi-Dastjerdi N, Soltany-Rezaee-Rad M, Sepehrizadeh Z, Roshandel G, \n421 Ebrahimifard F, Setayesh N. Genome expression analysis by suppression subtractive \n422 hybridization identified overexpression of Humanin, a target gene in gastric cancer \n423 chemoresistance. DARU J Pharm Sci. 2014;22(1):1-7. doi: 10.1186/2008-2231-22-14.\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n25\n424 18. Mottaghi-Dastjerdi N, Soltany-Rezaee-Rad M, Sepehrizadeh Z, Roshandel G, \n425 Ebrahimifard F, Setayesh N. Identification of novel genes involved in gastric carcinogenesis by \n426 suppression subtractive hybridization. Hum Exp Toxicol. 2015;34(1):3-11. doi: \n427 10.1177/0960327114532386.\n428 19. Li M, Li D, Tang Y, Wu F, Wang J. CytoCluster: a cytoscape plugin for cluster analysis \n429 and visualization of biological networks. Int J Mol Sci. 2017;18(9):1880. doi: \n430 10.3390/ijms18091880.\n431 20. Li Z, Lu T, Chen Z, Yu X, Wang L, Shen G, et al. HOXA11 promotes lymphatic \n432 metastasis of gastric cancer via transcriptional activation of TGFβ1. iScience. \n433 2023;26(8):107346. doi: 10.1016/j.isci.2023.107346.\n434 21. Wang Y, Liu H, Zhang M, Xu J, Zheng L, Liu P, et al. Epigenetic reprogramming in \n435 gastrointestinal cancer: biology and translational perspectives. MedComm. 2024;5(9):e670. doi: \n436 10.1002/mco2.670.\n437 22. Chen Q, Wang Y, Liu Y, Xi B. ESRRG, ATP4A, and ATP4B as Diagnostic Biomarkers \n438 for Gastric Cancer: A Bioinformatic Analysis Based on Machine Learning. Front Physiol. \n439 2022;13:905523. doi: 10.3389/fphys.2022.905523.\n440 23. Akhtar A, Hameed Y, Ejaz S, Abdullah I. Identification of gastric cancer biomarkers \n441 through in-silico analysis of microarray based datasets. Biochem Biophys Rep. 2024;40:101880. \n442 doi: 10.1016/j.bbrep.2024.101880.\n443 24. Cheng M, Hajime O, Yan S, Quan Y, Caroline Nadia F, Wu Y, et al. The role of \n444 adiponectin in gastric cancer. J Cancer Metastasis Treat. 2023;9:33. doi: 10.20517/2394-\n445 4722.2023.23.\n446 25. Sun Y, Nie W, Xiahou Z, Wang X, Liu W, Liu Z, et al. Integrative single-cell and spatial \n447 transcriptomics uncover ELK4-mediated mechanisms in NDUFAB1+ tumor cells driving gastric \n448 cancer progression, metabolic reprogramming, and immune evasion. Front Immunol. \n449 2025;16:1591123. doi: 10.3389/fimmu.2025.1591123.\n450 26. Chen H, Jing C, Shang L, Zhu X, Zhang R, Liu Y, et al. Molecular characterization and \n451 clinical relevance of metabolic signature subtypes in gastric cancer. Cell Rep. \n452 2024;43(7):114424. doi: 10.1016/j.celrep.2024.114424.\n453 27. Lau DK, Collin JP, Mariadason JM. Clinical developments and challenges in treating \n454 FGFR2-driven gastric cancer. Biomedicines. 2024;12(5):1117. doi: \n455 10.3390/biomedicines12051117.\n456 28. Edirisinghe O, Ternier G, Alraawi Z, Suresh Kumar TK. Decoding FGF/FGFR \n457 Signaling: Insights into Biological Functions and Disease Relevance. Biomolecules. \n458 2024;14(12):1622. doi: 10.3390/biom14121622.\n459 29. Cai X, Yang J, Guo Y, Yu Y, Zheng C, Dai X. Re-analysis of single cell and spatial \n460 transcriptomics data reveals B cell landscape in gastric cancer microenvironment and its \n461 potential crosstalk with tumor cells for clinical prognosis. J Transl Med. 2024;22(1):807. doi: \n462 10.1186/s12967-024-05606-9.\n463\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n26\n464 Figure Legends\n465 Fig 1. PPI network of the top 200 DEGs, with nodes colored by expression (red: upregulated, \n466 blue: downregulated). Constructed using STRING in Cytoscape (confidence score: 0.15).\n467 Fig 2. Functional enrichment analysis of upregulated genes. Top GO terms (BP, CC, MF) and \n468 Reactome pathways enriched among 85 upregulated GC genes, ranked by -Log 10FDR using \n469 ENRICHR. \n470 Fig 3. Functional enrichment of downregulated genes. Top GO terms and Reactome pathways for \n471 94 downregulated GC genes, ranked by -Log 10FDR using ENRICHR.\n472 Fig 4. Heatmap of Reactome pathway enrichment across top-ranked clusters. The presence of a \n473 pathway in a given cluster is indicated in blue, and absence is shown in white. Data are based on \n474 the top 20 enriched Reactome pathways for each of the top four PPI network clusters identified \n475 using the IPCA algorithm in CytoCluster. \n476 Fig 5. Grade-specific expression of 13 candidate genes in GC. Boxplots show transcript per \n477 million (TPM) values across normal and cancer grades 1-3 (TCGA data via UALCAN). \n478 Fig 6. Transcript expression levels of selected genes in GC patients based on nodal metastasis \n479 status. \n480 Fig 7. Stage-specific expression of 13 candidate genes in GC. Boxplots show transcript per \n481 million (TPM) values across normal tissue and cancer stages 1-4 (TCGA data via UALCAN).\n482 Fig 8. Survival analysis of ADIPOQ in GC patients. \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 19, 2026. ; https://doi.org/10.64898/2026.02.18.706548doi: bioRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}