Prognostic Model Identification of Ribosome Biogenesis-Related Genes in Pancreatic Cancer Based on Multiple Machine Learning Analyses

Yuan Sun

MS¹⁺

Yan Li 1

Anlan Zhang

MS²

Tao Hu

MS²

Ming Li 1✉ Email: liming@hospital.cqmu.edu.cn

1 Department of Hepatobiliary Surgery The First Affiliated Hospital of Chongqing Medical University Chongqing China

Yuan Sun, MS¹⁺, Yan Li, MS¹⁺,Anlan Zhang, MS², Tao Hu, MS², Ming Li, MD1*

Department of Hepatobiliary Surgery, The First Affiliated Hospital of Chongqing Medical University, Chongqing, China.

* Corresponding Author: Ming Li

Email: liming@hospital.cqmu.edu.cn

Abstract

Background

Pancreatic cancer is one of the most aggressive and lethal malignancies within the digestive system, characterized by an extremely low five-year survival rate and formidable treatment challenges. Ribosome biogenesis is a core process in protein synthesis, essential for cell growth, proliferation, and survival. Tumors typically exhibit abnormally high proliferation rates, with ribosome biogenesis often being excessively activated in cancer cells to synthesize a significant quantity of proteins required for tumor development. In pancreatic cancer, enhanced ribosome biogenesis may be associated with tumor drug resistance and malignant phenotypes, presenting it as a potential therapeutic target. Therefore, exploring the molecular mechanisms of ribosome biogenesis in pancreatic cancer may uncover new biomarkers and potential therapeutic targets, facilitating the development of personalized treatment strategies.

Methods

Ribosome biogenesis-related gene signatures were obtained from the TCGA database and Gene Cards. Multiple machine learning methods were employed to screen prognostic gene sets closely related to survival, followed by the construction of a risk model. The GEO database was utilized for external validation of the model. The GSE155698 single-cell RNA sequencing dataset was analyzed to evaluate gene expression profiles and associated module scores.

Results

A total of 60 ribosome biogenesis-related prognostic genes in pancreatic cancer were identified. Through univariate Cox regression and various machine learning techniques, nine key prognostic genes (ECT2, CKB, HMGA2, TPX2, ERBB3, SLC2A1, KRT13, PRSS3, CRABP2) were selected as highly sensitive and specific biomarkers for the diagnosis and prognosis of PAAD patients. These genes demonstrated high sensitivity and specificity as biomarkers for the diagnosis and prognosis of pancreatic cancer patients. The risk score feature constructed based on machine learning was found to be an independent factor associated with high mortality, advanced clinical pathological features, and chemotherapy resistance. Additionally, this feature was closely related to cell proliferation and cancer progression-related pathways, suggesting its potential role in promoting pancreatic cancer development through enhanced ribosomal division, mitosis-related processes, and microtubule cytoskeleton organization. Furthermore, the feature was closely associated with immunosuppressive cell infiltration and immune checkpoint expression, indicating its role as a key immunosuppressive factor within the tumor microenvironment. Single-cell RNA sequencing revealed higher expression of CKB, SLC2A1, ERBB3, CRABP2, and PRSS3 in pancreatic ductal epithelial cells.

Conclusions

This study provides insights into the potential connections between ribosome biogenesis-related molecular characteristics and clinical features, the tumor microenvironment, and clinical drug responses. The research underscores the critical role of ribosome biogenesis in the progression and treatment resistance of pancreatic cancer, offering valuable new perspectives for prognostic evaluation and therapeutic response prediction in pancreatic cancer.

Keywords

Pancreatic cancer

Ribosome Biogenesis

Prognosis

Tumor Microenvironment

Drug Sensitivity

1. Introduction

Pancreatic cancer is the most aggressive malignant tumor in the digestive system and ranks seventh in global cancer-related deaths. Due to the lack of early symptoms and effective diagnostic markers, most patients are diagnosed at an advanced stage, with a five-year survival rate of only approximately 9% [1]. Pancreatic ductal adenocarcinoma (PDAC) is the most common subtype, accounting for over 90% of pancreatic malignancies, and has a very poor prognosis. Although surgical resection remains the preferred treatment for PDAC, only 15–20% of patients are eligible for surgery at the time of diagnosis [2–3]. In recent years, new strategies such as FOLFIRINOX combined chemotherapy and neoadjuvant therapy have significantly improved the survival rates of some patients [4]. However, the efficacy of current chemotherapy regimens is limited by drug resistance [5], highlighting the urgent need for new adjuvant treatment strategies and biomarkers to better assess prognosis and guide personalized treatment.

Ribosome biogenesis is the process of assembling ribosomal complexes, a critical and complex process closely related to cell growth and protein synthesis, and is strictly regulated by numerous auxiliary factors [6–7]. In tumor cells, ribosome biogenesis is often upregulated and regulated by various signaling pathways, such as mTOR and Myc, to meet the high demand for protein synthesis required for rapid tumor growth, proliferation, and migration. This upregulation involves enhanced function of the nucleolus (the primary site of ribosomal RNA synthesis) and the overexpression of genes associated with ribosome assembly [8–10]. In pancreatic cancer research, ribosome biogenesis plays a significant role in the initiation and progression of the disease. For example, it has been reported that oncogenic RAS signaling enhances the RNA-binding activity of the nucleolar protein nucleolin (Ncl), reshaping the RNA-binding protein network of pancreatic cancer cells and leading to the phosphorylation of Ncl downstream of Erk1/2. The phosphorylation of Ncl enhances the binding of precursor rRNA and ribosome biogenesis. This nucleolin-dependent enhancement of ribosome biogenesis is crucial for RAS-induced proliferation of pancreatic cancer cells. Targeted therapies directed at this process can effectively inhibit tumor growth. [11–12]. Additionally, studies have shown that proteins involved in ribosome assembly, such as RRP9, can activate certain signaling pathways, thereby increasing the resistance of pancreatic cancer cells to the drug gemcitabine [13]. Moreover, ribosome biogenesis inhibitors targeting ribosome biogenesis may prevent the progression of pancreatic cancer by inducing oncogene-induced senescence in combination with ERK hyperactivation [14].

Therefore, this study aims to leverage multiple machine learning approaches to construct a robust prognostic model for assessing the clinical outcomes of pancreatic cancer patients. Additionally, it seeks to elucidate the critical role of ribosome biogenesis in pancreatic cancer initiation and progression, analyze somatic mutations associated with the disease, and identify novel therapeutic targets to advance treatment strategies.

2.Materials and methods

2.1 Data Sources

A total of 177 pancreatic cancer tissue samples (after excluding one sample with missing survival data) were selected from The Cancer Genome Atlas (TCGA) database. Clinical and pathological information, including age, sex, tumor grade, tumor stage, tumor type, tumor location, TNM staging, survival days, and survival status, were obtained. Transcriptomic sequencing data (seq-tpm and seq-count), methylation level data and microRNA expression matrix data were also acquired, along with transcriptomic sequencing data (seq-tpm and seq-count) from 180 normal pancreatic tissue samples sourced from the Genotype-Tissue Expression (GTEx) database. All transcriptomic sequencing data were normalized. Ribosome biogenesis-related genes were obtained from the GeneCards website. Validation cohorts, including ICGC-PACA-AU, GSE78229, GSE183795, and GSE62452, were retrieved from the International Cancer Genome Consortium (ICGC) and Gene Expression Omnibus (GEO) databases. Additionally, single-cell RNA sequencing data (scRNA-seq) for 10 PAAD tumor tissues (GSE155698) were used for single-cell analysis. Immunohistochemistry (IHC) data were downloaded from the Human Protein Atlas (HPA, http://www.proteinatlas.org) to verify protein expression in PAAD tissues.

2.2 Methods

2.2.1 Differential Gene Expression Analysis Between PAAD and Non-Tumor Tissues

To identify transcriptomic differences between PAAD and non-tumor tissues, 177 PAAD tumor samples were downloaded from the TCGA database (https://www.cancer.gov/ccg/research/genome-sequencing/tcga), and 180 normal pancreatic tissue samples were obtained from the GTEx database (https://gtexportal.org/home/) as controls. Using the DESeq2 package in R, differentially expressed genes (DEGs) between tumor and non-tumor samples were identified based on |fold-change| > 1.5 and adjusted p-value < 0.05. The results were visualized using the “ggplot2 “package.

2.2.2 Weighted Gene Co-expression Network Analysis (WGCNA) and Identification of Ribosome Biogenesis-Related Genes in PAAD

Weighted Gene Co-expression Network Analysis (WGCNA) was applied to further identify gene modules closely related to tumors in the TCGA-PAAD dataset [90]. Using the WGCNA package in R, differential expression genes (DEGs) of pancreatic cancer (PAAD) were analyzed. A gene expression matrix was constructed based on 4,725 DEGs from 357 samples for subsequent analysis. A soft-thresholding power of 5 was set to construct the gene co-expression network, calculating gene-gene co-expression similarity and adjacency, and converting it into a Topological Overlap Matrix (TOM). Hierarchical clustering based on TOM was then used to assign genes to modules. Through this process, gene modules significantly associated with clinical features were identified. In module analysis, key genes were screened by comparing gene significance (GS, indicating the correlation between the gene and clinical features) with module membership (MM, indicating the correlation between the gene and the module). Ultimately, upregulated genes associated with ribosome biogenesis (up CARGs) were extracted from these tumor-related modules. Simultaneously, 1,250 ribosome biogenesis-related genes were collected from the GeneCards database (https://www.genecards.org/, accessed on September 22, 2024). Using the VennDiagram package in R, an intersection analysis was performed between upregulated CARGs in PAAD and ribosome biogenesis-related genes, resulting in 60 significantly upregulated ribosome biogenesis-related genes in PAAD, which were prioritized for subsequent analyses.

2.2.3 Construction and Validation of Prognostic Signatures Based on Multiple Machine Learning Methods for Ribosome Biogenesis-Related Genes

To construct a prognostic model based on the identified 60 Ribosome Biogenesis-related genes (CARGs). Univariate Cox regression analysis was then performed, followed by the application of multiple machine learning methods using the R package Mime1. These methods included StepCox (stepwise regression), RSF (random survival forest), SuperPC (supervised principal component analysis), Lasso (least absolute shrinkage and selection operator), Survival-SVM (survival support vector machine), GBM (gradient boosting machine), CoxBoost (boosting algorithm for Cox regression), Enet (elastic net regression), Ridge (ridge regression), and plsRcox (partial least squares regression for Cox models). Models were constructed using individual and combined machine learning algorithms, and the concordance index (C-index) for each model was calculated. Furthermore, the predictive performance of the models was assessed using the TCGA-PAAD dataset as the training set, while the ICGC-PACA-AU, GSE78229, GSE183795, and GSE62452 datasets were used as independent validation sets. The average C-index across these validation datasets was also computed to ensure robustness.

Based on the median risk score calculated from the best model combination, the TCGA-PAAD training cohort was divided into low-risk and high-risk groups. Kaplan-Meier survival curves were plotted using the survival and survminer packages, and log-rank tests were performed to assess survival differences between the two groups. Receiver operating characteristic (ROC) curves were generated using the timeROC package to calculate the area under the curve (AUC), evaluating the model's sensitivity and specificity in predicting 2-year, 3-year, and 5-year survival probabilities. Additionally, the robustness and validity of the CARGs prognostic model were validated using GEO datasets. We also analyzed the correlation between risk scores and corresponding clinical indicators using the Wilcoxon rank-sum test and visualized the results with heatmaps generated by the heatmap package.

We applied univariate and multivariate Cox proportional hazards models to evaluate whether the CARGs prognostic model could independently predict outcomes after adjusting for other clinical and pathological features (such as age, sex, histological grade, pathological stage, and TNM stage). Based on univariate analysis, independent risk factors were integrated to construct a nomogram for predicting patient prognosis. Calibration curves were plotted to assess the consistency between predicted survival rates and actual observations.

2.2.4 Identification of Differentially Expressed Genes Between High and Low-Risk Groups and Functional Enrichment Analysis

Using the DESeq2 package in R, differentially expressed genes (DEGs) between the high-risk and low-risk groups (DEGs 2) were identified based on |fold-change| > 1 and adjusted p-value < 0.05. Functional enrichment analysis of the DEGs was performed using the clusterProfiler package to evaluate potential biological functions. Gene Set Enrichment Analysis (GSEA) was conducted using the reference gene sets “c2.cp.reactome.v7.0.entrez.gmt” and “h.all.v7.0.symbols.gmt”. Pathways were considered statistically significant if the normalized p-value < 0.05 and the false discovery rate (FDR) q-value < 0.25. The most enriched pathways were ranked based on the normalized enrichment score (NES). The limma package (version 3.54.2) was used to quantify the differential pathway activities between the high-risk and low-risk groups.

2.2.5 Tumor Microenvironment and Immune Infiltration Analysis

Firstly, we applied the single-sample Gene Set Enrichment Analysis (ssGSEA) algorithm to convert the gene expression data from the TCGA dataset into the abundance of 23 immune cell types. The proportions of infiltrating immune subtypes, including activated B cells, activated CD4⁺ T cells, M1 macrophages, and plasma cells, were assessed. Based on the differential analysis results, the infiltration levels of these immune cells were compared with the Ribosome Biogenesis risk scores. The Wilcoxon rank-sum test and Spearman correlation analysis were employed to further explore the relationship between risk scores and immune checkpoint expression. Additionally, the CIBERSORT algorithm was utilized to perform deconvolution analysis of the tumor immune microenvironment, calculating the infiltration levels of 22 immune cell types in each sample. Spearman correlation analysis was conducted to evaluate the correlation between risk scores and the infiltration levels of these 22 immune cell types. Furthermore, Spearman correlation analysis was performed to examine the relationship between the expression levels of the nine key genes and the infiltration degrees of the 22 immune cell types. Subsequently, the ESTIMATE algorithm was applied to evaluate immune and stromal infiltration scores and their correlation with Ribosome Biogenesis risk scores. The results were visualized using scatter plots and linear regression models. The Tumor Immune Dysfunction and Exclusion (TIDE) algorithm was employed to assess the response to immunotherapy, calculating immunotherapy scores for each sample. Additionally, immune scores under different immune checkpoint inhibitors (such as CD44 and TNFSF9) were compared between the high-risk and low-risk groups using immunotherapy data from the TCIA database. The results were visualized using box plots, and the Wilcoxon rank-sum test was used to assess the significance of the differences.

2.2.6 Mutation Analysis

We utilized the maftools package to analyze and visualize somatic mutation data between the high-risk and low-risk groups in the TCGA dataset, comparing the number and mutation spectra between the two groups. The maftools, remodeling2, and ggpubr packages were used to generate waterfall plots illustrating the mutation distribution of key mutated genes in each group. Additionally, we combined risk scores with tumor mutation burden (TMB) data to compare TMB differences between the high-risk and low-risk groups. Box plots displaying TMB differences were generated using the ggpubr package, and statistical tests were performed. Based on the median TMB value, samples were further classified into high-TMB and low-TMB groups. The chi-square test was used to evaluate the proportion differences between the high-risk and low-risk groups, and the results were visualized using bar charts.

2.2.7 Drug Sensitivity Evaluation

We employed the oncoPredict package in R to assess the predictive capability of Ribosome Biogenesis risk scores for 198 chemotherapy and targeted therapy drugs. In the TCGA-PAAD dataset (n = 177), we calculated the IC50 (half-maximal inhibitory concentration) for various drugs for each patient and categorized the patients into high-risk (n = 88) and low-risk (n = 89) groups. The Wilcoxon rank-sum test was used to analyze the differences in IC50 Z-scores between the high-risk and low-risk groups, exploring the potential value of Ribosome Biogenesis risk scores in predicting drug sensitivity.

2.2.8 Correlation Analysis of DNA Methylation Levels and microRNA with Key Genes and Risk Groups

Gene methylation levels and miRNA expression data for PAAD patients were obtained from the TCGA database. Differential analysis was conducted using the limma package in R, with the criteria of |fold-change| > 1 and adjusted p-value < 0.05. Correlation coefficients between the nine key genes and their corresponding gene methylation levels, as well as differentially expressed miRNAs, were calculated.

2.2.9 Single-Cell RNA Sequencing Analysis and HPA Immunohistochemistry Validation

Single-cell RNA sequencing (scRNA-seq) data, including raw sequencing data from 10 PAAD tumor tissues(GSE155698,https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE155698), were downloaded from the GEO database. Data processing and analysis were performed using Seurat (v4.3.0.1). Cells with a gene count (nFeature_RNA) greater than 200 but less than 5000, and UMI counts (nCount_RNA) between 200 and 50,000 were selected. Cells with mitochondrial UMI counts (pMT) exceeding 20% and red blood cell UMI counts (pHB) exceeding 5% were removed. Data normalization was performed using the “NormalizeData” and ScaleData functions, and 2000 highly variable genes were identified using the “FindVariableFeatures” function. Dimensionality reduction was conducted using Principal Component Analysis (PCA), and batch effects were corrected using the “RunHarmony” function. Cell clusters were visualized using t-SNE and UMAP, and cell clusters were manually annotated based on known marker genes. Additionally, immunohistochemistry (IHC) data were downloaded from the Human Protein Atlas (HPA, http://www.proteinatlas.org) to evaluate the high and low expression levels of the nine key genes in pancreatic cancer tissues.

Statistical Analysis

All statistical analyses were performed using RStudio (version R 4.3.2) and relevant R packages. All p-values were based on two-sided tests, and p-values less than 0.05 were considered statistically significant.

3. Results

3.1 Identification of Differentially Expressed Genes in Pancreatic Cancer and Ribosome Biogenesis-Related Genes Identification and Enrichment Analysis

In the TCGA and GTEx cohorts, a total of 4,625 differentially expressed genes (DEGs) were identified between PAAD samples and normal pancreatic tissues, meeting the criteria of p-value < 0.05 and absolute log₂ fold change > 1.5. Among these, 3,701 genes were upregulated, and 924 genes were downregulated (Fig. 1).

Fig. 1

identification of differentially expressed genes (DEGs) between PAAD tumor and normal tissues

3.2 Identification of Highly Correlated Gene Modules in TCGA-PAAD and Ribosome Biogenesis-Related Genes Identification and Enrichment Analysis

To further uncover key genes associated with tumors, Weighted Gene Co-expression Network Analysis (WGCNA) was employed to identify gene modules highly correlated with tumors in the TCGA-PAAD dataset. Initially, nine outlier samples were identified and excluded from the analysis through sample clustering (Fig. 2a). Subsequently, a soft-thresholding power of 5 was set (Fig. 2a), and a gene dendrogram was constructed to partition the genes into ten modules: red, purple, pink, black, yellow, magenta, green, turquoise, brown, and blue (Fig. 2b). The heatmap illustrating the correlations between different modules is also presented. The results demonstrated that tumor occurrence was significantly positively correlated with the turquoise module (correlation coefficient = 0.98, p-value = 6e-247) and the blue module (correlation coefficient = 0.88, p-value = 5e-115) (Fig. 2d). Additionally, the gene significance (GS, indicating the correlation between genes and clinical features) and module membership (MM, indicating the correlation between genes and modules) within the turquoise and blue modules were highly correlated, suggesting a strong association of genes in these modules with tumors (Fig. 2c). Genes in the yellow and green modules also exhibited strong correlations with tumors, whereas genes in other modules showed weaker associations (Figures S1). Finally, 2,875 upregulated genes (up CARGs) were extracted from the turquoise and blue modules and intersected with 2,259 ribosome biogenesis-related genes obtained from the GeneCards database using the VennDiagram package in R. This intersection analysis resulted in 60 key ribosome biogenesis-related genes (Fig. 3a). Principal Component Analysis (PCA) further validated that the expression of these 60 CARGs effectively distinguished tumor samples from non-tumor samples (Figures S2a). Additionally, a heatmap illustrated the differential expression of these 60 CARGs between tumor and non-tumor tissues (Figs. 3b).

Fig. 2

Identification of the highly correlated gene modules in WGCNA. (a) Soft-thresholding power determination: the left panel’s y-axis indicates the scale-free fit index, while the right panel’s y-axis indicates mean connectivity. The x-axis represents power values. (b) Dendrogram of differentially expressed genes clustered using a dissimilarity measure (1-TOM), with colors indicating identified gene modules. (c) Gene correlation scatter plots for the turquoise and blue modules: the x-axis represents module membership (MM), indicating the correlation between genes and modules, and the y-axis represents gene significance. (d) Correlations between gene modules and clinical traits: Boolean variables denote clinical phenotypes, representing "Normal" and "Tumor."

Gene Ontology (GO) enrichment analysis was performed on these 60 CARGs to explore potential molecular mechanisms associated with PAAD progression. The results indicated that these DEGs were primarily enriched in chromosomal regions, nuclear division, and mitosis-related processes, suggesting that ribosome biogenesis contributes to the synthesis of new proteins that may play roles in cell cycle control and cell division. In terms of cellular components, enrichment analysis revealed that these genes were closely associated with structures such as the spindle, blood microparticles, and the cell cortex, further indicating that the enrichment of ribosome biogenesis in these cellular components may be related to the enhanced proliferation and migratory capabilities of pancreatic cancer. Additionally, the enriched molecular functions were closely related to DNA binding, bending, and other related activities (Fig. 3c). These results collectively suggest that ribosome biogenesis may play a crucial role in PAAD development and progression by regulating the structure and function of genetic material, as well as extracellular signal transduction. Overall, these findings indicate that ribosome biogenesis may influence PAAD development and progression through the regulation of genetic material structure and gene expression.

Fig. 3

(a) dentification of the intersection among ribosome biogenesis-related genes, upregulated differentially expressed genes (DEGs), and genes from the WGCNA-identified tumor-associated turquoise and blue modules. (b) Expression levels of Gene 55 in tumor versus normal tissues. (c) GO term analysis of genes identified in the Venn diagram intersection among upregulated DEGs, ribosome biogenesis-related genes, and turquoise and blue modules. (d) Forest plot showing prognosis-related genes in PAAD identified through univariate Cox regression analysis.

3.3 Construction and Validation of Prognostic Model for Ribosome Biogenesis-Related Genes

Using multiple machine learning methods, we constructed a risk prediction model based on ribosome biogenesis-related genes from the TCGA database and validated it using GEO datasets. Initially, univariate Cox regression analysis of the 60 identified CARGs revealed 28 prognostically significant genes (Fig. 3d). We then constructed models using both individual and combined machine learning algorithms and calculated the concordance index (C-index) for each model. The average C-index across three datasets was computed to evaluate the predictive performance of all models. The optimal model combination was found to be a combination of Lasso (Least Absolute Shrinkage and Selection Operator) and RSF (Random Survival Forest), and the C-index was 0.65 (Fig. 4a). This model was trained using 1,000 decision trees, with a minimum node sample size of 5 and the random selection of 3 variables at each split. Ultimately, nine genes were identified as significantly impacting survival outcomes: CKB, ECT2, HMGA2, TPX2, ERBB3, SLC2A1, KRT13, PRSS3, and CRABP2 (Figures S3).

Fig. S3

The importance of nine genes in the prognostic model.

Based on this algorithm, risk scores for each sample were calculated, and the median risk score of 40.41 was used as the threshold to classify PAAD patients into low-risk (n = 89) and high-risk (n = 88) groups. Scatter plots demonstrated a significant decline in overall survival (OS) with increasing individual risk scores (Fig. 4b). Kaplan-Meier survival curves and log-rank tests further confirmed that OS was significantly lower in the high-risk group compared to the low-risk group (p < 0.0001; Fig. 4b).

Fig. 4

Construction of a machine learning–based risk-scoring model for PAAD patients using the TCGA database. (a) C-index values for the test and validation prognostic signatures. (b) Patient risk score distribution and Kaplan–Meier survival analysis of TCGA-PAAD patients in high-risk versus low-risk groups. (c) ROC curve analysis of the model’s performance for predicting 2-, 3-, and 5-year overall survival (OS).

To evaluate the performance of the prognostic model, time-dependent ROC curve analysis was conducted to assess the sensitivity and specificity of the model. The area under the curve (AUC) for predicting 2-year, 3-year, and 5-year survival were 0.976, 0.987, and 0.938, respectively, indicating high accuracy in predicting OS (Fig. 4c). Subsequently, the feasibility of using ribosome biogenesis features to predict PAAD patient prognosis was validated using three GSE datasets and one ICGC dataset. Kaplan-Meier survival curves showed that patients in the low-risk group had better OS than those in the high-risk group (Figs. 5a, 5c, 5e, and 5g). Robust AUC values for the 2-year, 3-year, and 5-year survival predictions were also observed in both the training and validation cohorts (Figs. 5b, 5d, 5f, and 5h). Clinical correlation analysis revealed significant associations between risk scores and follow-up tumor status, radiation therapy, and residual tumors (Figs. 6a-d).In summary, the ribosome biogenesis-related prognostic features constructed using multiple machine learning methods demonstrated excellent performance in predicting survival outcomes for PAAD.

Fig. 5

Survival analysis results of the risk-scoring model for PAAD patients from the GEO and ICGC datasets. (a,c,e,g) Kaplan–Meier survival curves for GEO PAAD patients classified into high-risk and low-risk groups. (b,d,f,h) ROC curve analysis for 2-, 3-, and 5-year overall survival (OS). Abbreviations: OS, overall survival; ROC, receiver operating characteristic.

Fig. 6

(a) A complex heatmap depicting clinical characteristics in high- and low-risk groups.(b)Clinical features associated with the risk score, including tumor status at follow-up tumor status, radiation therapy, and residual tumor.

3.4 Ribosome Biogenesis-Related Prognostic Features as Independent Prognostic Factors in PAAD

To determine independent prognostic factors for PAAD patients, we included ribosome biogenesis-related risk scores and various clinical characteristics (including age, sex, histological grade, T stage, M stage, N stage, history of malignancy, follow-up tumor status, tumor location, DCC histological type, history of radiotherapy, residual tumor, and chronic pancreatitis) in univariate and multivariate Cox regression analyses. The univariate Cox regression results indicated that age (HR = 1.028, 95% CI: 1.007–1.049), histological stage (HR = 2.350, 95% CI: 1.078–5.126),N stage (HR = 2.0961, 95% CI: 1.249–3.519), follow-up tumor status (HR = 2.491, 95% CI: 1.077–5.762), radiation therapy (HR = 0.453, 95% CI: 0.257–0.798), DCC histological type (HR = 1.770, 95% CI: 1.077–2.908), and residual tumor (HR = 1.770, 95% CI: 1.097–2.857) were all significant prognostic factors in the TCGA cohort (Fig. 7a). In the multivariate Cox regression analysis, the risk score remained a significant independent prognostic factor closely associated with prognosis (Fig. 7b).

Figure 7.Univariate (a) and multivariate (b) regression analysis of the risk score in the TCGA-PAAD cohort; Cox proportional hazards regression.

Seven significant clinical variables, along with the risk score, were incorporated into the model. Based on the stepwise Cox regression model, a nomogram was constructed to predict 2-year, 3-year, and 5-year OS. The parameters included in the nomogram were risk score, age, N stage, histological grade, follow-up tumor status, history of radiotherapy, DCC histological type, and residual tumor. The results emphasized the predictive capability of ribosome biogenesis-related risk scores in PAAD prognosis (Fig. 8a). The AUC values for predicting 2-year, 3-year, and 5-year OS were 0.986, 0.990, and 0.949, respectively (Fig. 8b). Subsequently, calibration curves were plotted to assess the performance of the nomogram model. The calibration curves closely aligned with the 45-degree diagonal, indicating a high consistency between the predicted survival probabilities and the actual observed outcomes in the TCGA-PAAD cohort (Fig. 8c).

Fig. 8

(a) This nomogram estimates 2-year, 3-year, and 5-year overall survival probabilities for patients with pancreatic adenocarcinoma. The included predictors are risk score, age, stage, AJCC pathological N, follow-up tumor status, WHO histological classification, radiation therapy, and residual tumor status. (b) AUC values of the nomogram for predicting 2-year, 3-year, and 5-year survival. (c) Calibration curves showing the agreement between the nomogram’s predicted probabilities and the actual observed outcomes at 2-year, 3-year, and 5-year intervals. The x-axis represents predicted survival probability, and the y-axis represents observed survival probability.

3.5 Functional Analysis of Ribosome Biogenesis-Related Gene Set and Mutation Analysis

To investigate the biological behaviors associated with different Ribosome Biogenesis risk score groups, we employed the GSVA package in R for functional annotation of TCGA-PAAD samples. The heatmap of GSVA-KEGG results revealed that taurine and hypotaurine metabolism-related pathways were significantly enriched in the high-risk group. Additionally, pathways related to the cell cycle and DNA replication were also markedly enriched in the high-risk group. Furthermore, the p53 pathway exhibited higher activity in the high-risk group. In contrast, protein degradation pathways were significantly enriched in the low-risk group (Fig. 9c).

Simultaneously, GSVA-HALLMARK enrichment analysis demonstrated that multiple pathways associated with cancer progression and cell proliferation were significantly enriched in the high-risk group. These pathways include MYC target-related pathways, DNA repair, G2/M checkpoint-related pathways, mTORC1 signaling, hypoxia, and apoptosis-related pathways. Notably, cancer-related pathways such as the p53 pathway, Notch signaling pathway, and TGF-β signaling were more active in the high-risk group compared to the low-risk group (Figs. 9a–b), which was generally consistent with the GSVA-KEGG results. Gene Set Enrichment Analysis (GSEA) further indicated significant enrichment of pathways related to the regulation of cell cycle checkpoints, mitosis, and keratinocyte envelope formation (Fig. 9e).

To further explore the differential gene expression between the CARGs and the risk scores in PAAD patients, we identified DEGs (DEGs 2) between the high-risk and low-risk groups and performed GO-KEGG enrichment analysis using the “org.Hs.eg.db” and “clusterProfiler” packages. The results revealed that these DEGs were enriched in various cellular components, including neuronal cell bodies, synaptic membranes, transmembrane transporter complexes, and ion channel complexes related to neuronal cell membrane structures and channel complexes. In biological processes, these genes were primarily enriched in processes such as the regulation of chemical synaptic transmission, membrane potential regulation, and potassium ion transport. In terms of molecular functions, significant enrichment was observed in potassium and metal ion transmembrane transporter activities, voltage-gated channel activities, and ion channel activities (Fig. 9d). Additionally, KEGG pathway analysis indicated enrichment in neuroactive ligand-receptor interactions and dopaminergic synapse-related pathways (Fig. 9f), suggesting that neuroregulation and metabolic abnormalities may play crucial roles in the initiation and progression of PAAD.

Fig. 9

enrichment analysis. (a) The top 20 significantly enriched GSVA-HALLMARK pathways in both high- and low-risk groups. (b) The top 20 GSVA-HALLMARK pathways showing increased enrichment in the high-risk group. (c) The top 20 significantly enriched GSVA pathways in the high- and low-risk groups. (d) The most significant and commonly enriched GO terms in the TCGA dataset. (e) The important pathway of GSEA analysis.(f) The most significant and commonly enriched KEGG pathways in the TCGA dataset.

We also analyzed the genomic mutation characteristics associated with the CARGs gene set, including mutation counts and mutation spectra. Utilizing the maftools package, we visualized the mutation patterns of 168 patients from the TCGA-PAAD cohort. The results indicated that the overall mutation rate was higher in the high-risk group compared to the low-risk group, suggesting greater genomic instability in the high-risk group. TP53 and KRAS were the most frequently mutated genes in the high-risk group, with mutation rates significantly higher than those in the low-risk group (TP53: 70% vs. 51%, KRAS: 80% vs. 50%) (Figs. 10a–b). These mutations are closely associated with cancer progression and poor prognosis across various cancers, indicating their potential critical roles in disease progression among high-risk patients. Additionally, CDKN2A and SMAD4 exhibited relatively high mutation frequencies in both groups, with slightly higher rates in the high-risk group, further supporting their importance in tumor suppression and signal transduction. Furthermore, TTN also displayed moderately high mutation frequencies in both groups, warranting further investigation.

Regarding tumor mutation burden (TMB), the proportion of high-TMB samples was higher in the low-risk group compared to the high-risk group, and this difference was statistically significant based on the chi-square test (Figs. 10c–d, p = 0.0032). This suggests that the high-risk group may be associated with poorer responses to immunotherapy.

Fig. 10

(a) The top 20 genes with the highest mutation frequencies in the high-risk group. (b) The top 20 genes with the highest mutation frequencies in the low-risk group. (c) A bar chart illustrating the distribution of tumor mutation burden (TMB) across the high- and low-risk groups. (d) TMB distribution differences between the two groups (Wilcoxon test).

3.6 Correlation Analysis of Immune Infiltration and Tumor Microenvironment

To explore the relationships among Ribosome Biogenesis, immune infiltration, and the tumor microenvironment, we first conducted single-sample Gene Set Enrichment Analysis (ssGSEA) to compare immune cell infiltration levels between the high-risk and low-risk groups. The results indicated that in the high-risk group, the infiltration levels of CD4⁺ T cells and Type II helper T cells were significantly higher than those in the low-risk group (P < 0.05). Conversely, the low-risk group exhibited significantly higher infiltration levels of activated B cells and eosinophils (P < 0.01), as well as higher infiltration levels of monocytes and macrophages (P < 0.05) (Fig. 11a).

In terms of immune functions, the high-risk group showed elevated levels of antigen-presenting cell co-inhibition, immunosuppressive inflammation, and Type I interferon responses compared to the low-risk group (P < 0.05) (Fig. 11b). Using the ESTIMATE algorithm, we examined the relationship between Ribosome Biogenesis and the tumor microenvironment in 177 PAAD patients from TCGA. ESTIMATE analysis revealed that the high-risk group had higher tumor purity compared to the low-risk group (Fig. 11f), while the ESTIMATE score and immune score were lower in the high-risk group, with no significant difference in the stromal score between the two groups (Fig. 11e).

In immune checkpoint analysis, CD44, LGALS9, CD160, IDO2, and TNFSF9 were significantly more highly expressed in the high-risk group (Fig. 11c), suggesting that immune checkpoint inhibitors (ICIs) targeting these molecules may offer potential therapeutic benefits. The Tumor Immune Dysfunction and Exclusion (TIDE) algorithm, a computational tool for predicting responses to immune checkpoint blockade (ICB), was employed to assess the potential clinical responses to ICB in different risk groups. The results indicated that the relationship between patient immune treatment responses and their risk scores did not reach statistical significance, suggesting that patients in the high- and low-risk groups may not differ significantly in their responses to conventional ICB treatments (Fig. 11f).

Fig. 11

TCGA tumor microenvironment between high-group and low-risk group. (a) Boxplots of the 23 immune signature ssGSEA scores comparing high-risk and low-risk patients in TCGA. (b) Boxplots of immune function ssGSEA scores comparing high-risk and low-risk patients in TCGA. (c) Differential expression of ADORA2A, BTLA, CD160, CD200, CD276, CD44, HHLA2, IDO2, KIR3DL1, LGALS9, and TNFSF9 between high- and low-risk groups in TCGA. Statistical significance is indicated as follows: *p < 0.05, **p < 0.01, ***p < 0.001, ns = not significant. (d) Spearman’s correlation coefficients between the 9 genes and infiltration levels of 23 immune cell types. (e) Differences in stromal score, ESTIMATE score, and immune score between the high-risk (n = 88) and low-risk (n = 89) groups in TCGA PAAD patients. (e) Differences in tumor purity between the high-risk (n = 88) and low-risk (n = 89) groups in TCGA PAAD patients. (g) Differences in TIDE scores between the high-risk (n = 88) and low-risk (n = 89) groups.

Subsequently, using the CIBERSORT algorithm, we performed deconvolution analysis of the tumor immune microenvironment and assessed the correlation between infiltration levels and risk scores. The results demonstrated that activated mast cells and activated B cells were negatively correlated with risk scores, while M1 macrophages and activated dendritic cells were positively correlated with risk scores, with all correlations being statistically significant (P < 0.05) (Figs. 12a–c). Spearman correlation analysis revealed that among the nine CARGs, the expression levels of ECT2 were strongly correlated with the infiltration levels of Type II helper T cells, ERBB3 and PRSS3 with the infiltration levels of Type 17 helper T cells, SLC2A1 with CD56⁺dim natural killer (NK) cells, and TPX2 with activated CD4⁺ T cells (Fig. 11d).

Fig. 12

(a) Boxplots depicting the CIBERSORT scores of 22 immune cells of the high-risk patients compared to low-risk patients in TCGA.(b)The Correlation Coeffcient of 22 immune cells of the high-risk patients with risk score.(c)Correlation of B_cells_naive、Dendritic_cells_activated、Macrophages_M1 and Mast_cells_activated with risk score.

3.7 Drug Sensitivity Prediction and CARGs Risk Score

We utilized cell line data from the Cancer Genome Project (CGP) database to predict the IC50 of commonly used chemotherapy drugs for different risk groups of PAAD patients. Gemcitabine is a first-line treatment for pancreatic cancer and can be used alone or in combination for PAAD chemotherapy [15]. According to our study, patients in the low-risk group exhibited significantly higher sensitivity to gemcitabine compared to those in the high-risk group (P < 0.001), with a correlation coefficient of 0.4 between risk scores and gemcitabine IC50 (Fig. 13a). This suggests that the median inhibitory concentration (IC50) for gemcitabine is significantly lower in the low-risk group, indicating that low-risk patients may be more sensitive to gemcitabine, while high-risk patients may be more prone to developing resistance.

Additionally, the IC50 values for other first-line chemotherapy drugs, such as platinum-based drugs (oxaliplatin, cisplatin) and topoisomerase inhibitors (irinotecan), were also significantly lower in the low-risk group (P < 0.001, Figs. 13b–d). Notably, high-risk patients exhibited significantly lower IC50 values for paclitaxel, erlotinib, Laptinib, and Acetalax compared to the low-risk group (Figs. 13e–f and Figures S4a–b), suggesting that these drugs may offer potential therapeutic benefits for patients in the high-risk category.

Fig. 13

(a) Correlation analysis of the Risk Score with Gemcitabine IC50 and Boxplots of the estimated IC50 of Gemcitabine in the high- (n = 88) and low-risk (n = 89) group. (b) Correlation analysis of the Risk Score with Cisplatin IC50 and Boxplots of the estimated IC50 of Cisplatin in the high- (n = 88) and low-risk (n = 89) group. (c) Correlation analysis of the Risk Score with Irinotecan IC50 and Boxplots of the estimated IC50 of Irinotecan in the high- (n = 88) and low-risk (n = 89) group. (d) Correlation analysis of the Risk Score with Oxaliplatin IC50 and Boxplots of the estimated IC50 of Oxaliplatin in the high- (n = 88) and low-risk (n = 89) group. (e) Correlation analysis of the Risk Score with Paclitaxel IC50 and Boxplots of the estimated IC50 of Paclitaxel in the high- (n = 88) and low-risk (n = 89) group. (f) Correlation analysis of the Risk Score with Erlotinib IC50 and Boxplots of the estimated IC50 of Erlotinib in the high- (n = 88) and low-risk (n = 89) group.

3.8 Analysis of Gene Methylation Levels and miRNA Expression

We conducted differential analysis of gene methylation levels and miRNA expression between the high-risk and low-risk groups. The results indicated that there were no significant differences in the methylation levels of corresponding genes between the high-risk and low-risk groups (Figures S5a). Furthermore, there was no apparent correlation between the methylation levels of the nine key prognostic CARGs and their mRNA expression levels (Figures S5b). This suggests that these genes may not be significantly associated with methylation levels.

Fig. S5

(a) differentially methylated genes between high-risk and low-risk groups. (b) Spearman’s correlation between the methylation levels and expression levels of the nine genes (c) Differentially expressed microRNA genes between high-risk and low-risk groups.(d) Spearman’s correlation between the expression levels of three differentially expressed microRNAs and the expression levels of the nine genes.

Differential analysis of miRNA expression between the high-risk and low-risk groups revealed that miR-203a-3p, hsa-miR-196b-5p, and hsa-miR-205-5p were highly expressed in the high-risk group (Figure S5c). However, the correlation between the mRNA expression levels of the nine genes and miRNA expression levels did not reach a correlation coefficient of 0.7 (Figure S5d).

3.9 Validation of Ribosome Biogenesis-Related Genes at the Single-Cell Level and Protein Expression Analysis

To further elucidate the specific roles of CARGs in PAAD progression, we performed single-cell RNA sequencing (scRNA-seq) analysis on 10 samples (GSE155698) to study the expression profiles of LLPS-related genes in the pancreatic malignant tumor microenvironment. After quality control filtering, we obtained 24,094 cells from 10 primary pancreatic tumors. These cells were subsequently merged, clustered, and annotated (Fig. 14a). Based on cell type-specific marker genes, the cells were classified into T cells, macrophages, endothelial cells, pancreatic ductal cells, NK cells, fibroblasts, B cells, and pancreatic cancer stem cells (Fig. 14b).

Next, we plotted the expression profiles of nine ribosome biogenesis-related genes, including ECT2, CKB, HMGA2, TPX2, ERBB3, SLC2A1, KRT13, PRSS3, and CRABP2. As shown in Fig. 14c, CKB, SLC2A1, ERBB3, and CRABP2 were predominantly highly expressed in pancreatic ductal cells, while PRSS3 was highly expressed in both pancreatic acinar and ductal cells. The other genes exhibited only minimal expression (Fig. 14c). To further quantify the overall expression levels of these genes across different cell types, we calculated a module score based on the nine genes. This module score represented the collective expression level of these nine genes within individual cells. The results indicated higher expression levels in ductal epithelial cells and acinar cells (Fig. 14d). These findings suggest that ribosome biogenesis may play critical roles in these cell types.

Fig. 14

Expression of nine ribosome biogenesis–related genes in a single-cell RNA sequencing dataset. (a) Cell cluster annotations. (b) Cell type annotations. (c) Expression patterns of ECT2, CKB, HMGA2, TPX2, ERBB3, SLC2A1, KRT13, PRSS3, and CRABP2. (d) Violin plots showing the expression levels of these ribosome biogenesis–related genes across various cell types in pancreatic cancer, as revealed by single-cell RNA sequencing analysis.

Immunohistochemistry (IHC) data were downloaded from the Human Protein Atlas (HPA, http://www.proteinatlas.org) to evaluate the expression levels of eight genes (CKB, HMGA2, TPX2, ERBB3, SLC2A1, KRT13, PRSS3, and CRABP2) in pancreatic cancer tissues. IHC analysis demonstrated varying degrees of protein expression, ranging from low to high, in representative PAAD samples (Figs. 15–16).

Fig. 15.16

Protein expression levels of ECT2, CKB, HMGA2, TPX2, ERBB3, SLC2A1, KRT13, PRSS3, and CRABP2 in pancreatic adenocarcinoma tumor tissues

3.10 Cell Communication

The intercellular signaling network diagram (Fig. 17a) reveals complex signal transduction relationships among various cell types, including macrophages, endothelial cells, pancreatic ductal cells, pancreatic acinar cells, and fibroblasts. In this network, node size represents different cell types, while the thickness of the connections indicates the strength of the signaling interactions.Heatmaps illustrating signaling patterns across cell types (Figs. 17b and 17c) demonstrate that high-scoring ductal epithelial cells exhibit strong outgoing signals in pathways such as MIF, MK, and PDGF. Conversely, macrophages show robust incoming signals in pathways including MIF, SPP1, and CSF. These findings highlight the diversity of intercellular communication.Collectively, these results elucidate the complexity of the intercellular communication network and the specificity and intensity of signaling pathways between distinct cell types. The heatmap depicting communication probability and strength between different cell types (Fig. 17d) indicates that high-scoring ductal epithelial cells engage in strong ligand-receptor interactions with B cells and macrophages. This visualization underscores the key communications among various cell types within the tumor microenvironment through diverse signaling pathways.Furthermore, we conducted an analysis of the MIF signaling pathway network (Figs. 17e and 17f), where nodes represent different cell types and the thickness of the connections reflects the intensity of MIF signal transduction. This analysis emphasizes the complex and critical role of MIF signaling in mediating intercellular communication.

Fig. 17

Single-cell communication analysis (a) network diagram illustrating intercellular signaling pathways (b) Heatmap showing outgoing signaling patterns across different cell types. (c) Heatmap displaying the incoming signaling pathways in various cell types. (d) Bubble plot illustrating the communication probability and strength between different cell types (e) MIF signaling pathway cell communication network diagram(circular layout). (f) MIF signaling pathway cell communication network diagram(hierarchical layout)

4. Discussion

Ribosome biogenesis is the highly precise process of constructing ribosomal particles, intricately linked to cellular protein synthesis, proliferation, differentiation, and apoptosis. This process entails the coordinated synthesis of four ribosomal RNAs (rRNAs), eighty ribosomal proteins (RPs), and seventy small nucleolar RNAs (snoRNAs) [16,17]. Within the nucleolus of eukaryotic cells, proteins associated with rRNA, RPs, and snoRNAs collaborate with RNA polymerases I and III to facilitate the generation and maturation of ribosomal precursors, specifically the pre-60S and pre-40S subunits. Following their export to the cytoplasm, ribosomes undergo final assembly and maturation [18].

In tumors, various oncogenic signaling pathways enhance ribosome biogenesis and protein synthesis, thereby promoting tumor growth and progression. The oncogene Myc serves as a pivotal regulator in cancer progression by directly activating rRNA synthesis, thereby increasing ribosome biogenesis rates. Nucleolin (Ncl) further promotes rRNA transcription and pre-rRNA processing, aiding oncogenesis. Additionally, Ncl interacts with telomerase reverse transcriptase (TERT) in the nucleoplasm, facilitating its localization to the nucleolus, preventing cancer cell senescence, and sustaining proliferation. Emerging evidence indicates that targeting ribosome biogenesis is a promising therapeutic strategy for cancer patients [19–23]. In pancreatic cancer, mutations in oncogenic RAS and various cancer signaling pathways are potentially linked to ribosome biogenesis [11–13]. Abnormalities in ribosomal proteins and RNA-binding proteins, such as heterogeneous nuclear ribonucleoproteins (hnRNPs) and RRP9, have been shown to promote cancer cell proliferation and chemotherapy resistance [13,24]. Furthermore, studies have demonstrated that inhibitors of ribosome biogenesis, like BMH-21, which targets RNA polymerase I to inhibit rRNA synthesis, exhibit significant anticancer effects in mouse models [25]. Consequently, ribosome biogenesis has increasingly become a central focus in pancreatic cancer research.

In our study, we employed multiple machine learning methods to identify nine ribosome biogenesis-related genes that are highly associated with prognosis: ECT2, CKB, HMGA2, TPX2, ERBB3, SLC2A1, KRT13, PRSS3, and CRABP2.

ECT2 is a RhoGTPase activator essential for cytokinesis. ECT2 has been identified as an oncogene when ectopically expressed in NIH/3T3 fibroblasts, with cytoplasmic and nuclear ECT2 exhibiting distinct oncogenic mechanisms [26]. Studies have shown that in cancers such as lung and ovarian cancer, cytoplasmic ECT2 interacts with the PKCi-Par6 complex and, through PKCi-mediated phosphorylation, drives the RAC1-MEK-ERK signaling axis essential for transformational growth in tumor [27–29]. In colorectal and lung cancers, nuclear ECT2 promotes rDNA transcription and ribosome biogenesis, which may be particularly important for more invasive and highly proliferative tumors [30–31]. In pancreatic cancer, research has demonstrated that YAP1 regulates the expression of ECT2, inducing malignant transformation. Nuclear ECT2 is upregulated during the transition from premalignant lesions to pancreatic ductal adenocarcinoma (PDAC), and high levels of YAP1 or ECT2 expression are associated with poorer overall survival in PDAC patients. ECT2 plays a crucial role in the growth and metastasis of pancreatic cancer [32].

CKB (Creatine Kinase B-type) primarily functions in cellular energy metabolism. In particular tumors, CKB may support rapid proliferation and the protein synthesis demands of ribosome biogenesis by maintaining high levels of energy metabolism [33]. Due to CKB’s primary role in cellular energy metabolism, its specificity in tumors is relatively limited. Currently, no studies have directly linked CKB to pancreatic cancer. However, its high expression may be associated with the high-energy metabolic characteristics and metabolic reprogramming of tumors, suggesting a need for further investigation.

High mobility group (HMG) proteins are a class of non-histone, chromatin-associated molecules involved in maintaining and regulating DNA functions, including replication, recombination, transcription, and DNA repair. High mobility group protein A2 (HMGA2) primarily regulates gene expression by binding to AT-rich regions of DNA and interacting with RNA-binding proteins, thereby influencing the ribosome biogenesis process [39–41]. Notably, HMGA2 is re-expressed in nearly all human malignancies and promotes tumorigenesis through various mechanisms [42]. In pancreatic cancer, HMGA2 has been extensively studied. Research indicates that HMGA2 is involved in maintaining epithelial-mesenchymal transition (EMT) in human pancreatic cancer cells and serves as a prognostic marker for metastatic cancer cell states in primary PDAC. However, HMGA2 does not significantly impact chemotherapy resistance. Various studies suggest that HMGA2 is a potential therapeutic target for pancreatic cancer treatment [43–44].

TPX2 is a critical regulator of cell division, primarily involved in spindle assembly and maintaining genomic stability. It regulates ribosome biogenesis during mitosis by controlling the cell cycle [45]. Overexpression of TPX2 in many malignancies, coupled with its essential role in mitotic regulation, suggests that TPX2 is a potential oncogenic driver [46]. In pancreatic cancer, studies have demonstrated that TPX2 expression is significantly elevated compared to normal tissues. Besides, targeting TPX2 with small interfering RNA effectively inhibits the proliferation of pancreatic cancer cells, and TPX2 knockout enhances the sensitivity of these cells to paclitaxel treatment. Thus, TPX2 emerges as a promising therapeutic target for pancreatic cancer [47].

ERBB3, a member of the epidermal growth factor receptor (EGFR) family and the tyrosine kinase receptor family, primarily functions through heterodimerization with ERBB2. This dimerization activates downstream signaling pathways, such as PI3K/AKT and RAS/RAF/MEK, promoting cell growth, proliferation, and survival [48]. In various cancers, ERBB3 has been found to localize to the nucleolus, interacting with nucleolar components to participate in cell survival and proliferation by regulating transcription factors involved in ribosome biogenesis [49–51]. In pancreatic cancer, dysregulated ErbB signaling plays a pivotal role in tumorigenesis. Studies have shown that NRG-1 derived from cancer-associated fibroblasts promotes PDAC tumorigenesis through the ErbB3-AKT signaling pathway, identifying the NRG-1/ErbB3 axis as an attractive molecular target to disrupt tumor-stroma epithelial interactions within the PDAC microenvironment [52]. Additionally, valproic acid treatment may induce apoptosis in pancreatic cancer cells by downregulating EGFR, ERBB2, and ERBB3 through microRNAs targeting the ErbB family members [53]. The ErbB signaling pathway has thus become a hotspot for targeted therapy in pancreatic cancer.

SLC2A1, also known as GLUT1, is a key gene involved in cellular glucose metabolism [54]. Overexpression of SLC2A1 further enhances glycolysis and cell proliferation in various cancers, promoting the protein synthesis required for ribosome biogenesis through energy metabolism [55]. In pancreatic cancer, studies have demonstrated that SLC2A1 promotes metabolic reprogramming and chemoresistance in PDAC through the GLUT1/ALDOB/G6PD axis [56]. Additionally, SLC2A1 is regulated by upstream factors such as FOXD1, which directly promotes SLC2A1 transcription and enhances GLUT1 expression by regulating aerobic glycolysis, ultimately promoting pancreatic cancer cell proliferation, invasion, and metastasis [57].

KRT13 (Keratin 13) is an intermediate filament protein predominantly expressed in epithelial cells. In cancer, the KRT13/PG/DSP complex interacts directly to alter PG expression and nuclear translocation, thereby regulating downstream c-Myc signaling. c-Myc enhances ribosome biogenesis and function by regulating multiple biological processes, supporting the high demand for protein synthesis in rapidly proliferating cancer cells, and increasing epithelial-to-mesenchymal transition and stem cell-like phenotypes [58–59]. In pancreatic cancer, KRT13 is upregulated in pancreatic cancer stem cell-like cells and is associated with radiotherapy resistance [60].

PRSS3 (Protease, Serine 3) is a member of the small serine protease family in the pancreas. Research indicates that PRSS3 plays a critical role in cancer progression, particularly in promoting tumor cell invasion and metastasis [61–62]. Specifically, in pancreatic cancer, PRSS3 is overexpressed in metastatic cell lines. It upregulates VEGF expression through the PAR1-mediated ERK pathway to promote angiogenesis. As a serine protease, PRSS3 enhances extracellular matrix (ECM) degradation, altering the tumor microenvironment and thereby promoting tumor progression and metastasis in vivo. PRSS3 plays an essential role in the progression, metastasis, and prognosis of human pancreatic cancer, making it a viable target for therapeutic intervention [63–64].

CRABP2 (Cellular Retinoic Acid Binding Protein 2) is a crucial intracellular retinoic acid transport protein that transports retinoic acid (RA) to the nucleus and interacts with its receptor complex, retinoic acid receptor (RAR). CRABP2 regulates cell proliferation, apoptosis, and metastasis by transporting RA to the nucleus and interacting with RAR [65–67]. High expression of CRABP2 is associated with increased invasiveness and poor prognosis in various cancers. Studies have shown that CRABP2 interacts with the integrin/FAK/ERK signaling pathway, enhancing cellular responses to growth factors, increasing cellular metabolic activity, and promoting ribosome biogenesis, thereby leading to tumor invasiveness and metastasis [68]. However, its role in pancreatic cancer remains unexplored, presenting a potential avenue for future research.

Consistent with the reports mentioned above, our results indicate that the features of these nine genes are closely related to the clinical characteristics of malignant tumors and chemotherapy resistance. Furthermore, this nine-gene signature risk model can independently predict overall survival outcomes beyond known clinical and pathological risk factors.

Our Analysis Centered on Multiple Machine Learning Algorithms to Differentiate High and Low-Risk Groups Based on Risk Scores and Identify Potential Drugs Significantly Affecting Disease Prognosis. High-risk PAAD patients appear to exhibit resistance to chemotherapy drugs that interfere with DNA synthesis, such as gemcitabine, oxaliplatin, and cisplatin, as well as topoisomerase I inhibitors like irinotecan. Mechanisms underlying resistance to DNA synthesis inhibitors generally include enhanced DNA repair capabilities and altered cellular metabolic pathways. Enhanced DNA repair can lead to the restoration of DNA damage induced by topoisomerase I inhibitors, thereby diminishing the cytotoxic effects of irinotecan [69–71]. These mechanisms may be potentially linked to our finding that high-risk groups are enriched in DNA repair pathways.

Additionally, we discovered that paclitaxel may offer benefits to high-risk patients. Paclitaxel inhibits cell mitosis by stabilizing microtubule structures, preventing cancer cells from completing the division cycle [72]. The nanoparticle form of paclitaxel is often used in combination with gemcitabine as a first-line chemotherapy regimen for pancreatic cancer and metastatic pancreatic cancer [73]. This aligns with our enrichment results, where high-risk group patients are more enriched in pathways regulating cell cycle checkpoints and mitosis, potentially serving as potential targets for paclitaxel.

Erlotinib and Acetalax are both EGFR tyrosine kinase inhibitors (TKIs). Erlotinib, in combination with gemcitabine, remains a first-line chemotherapy option for advanced pancreatic cancer. However, compared to the FOLFIRINOX regimen, it has limitations and is suitable for specific patient populations who cannot tolerate multiple chemotherapy regimens [74]. In our cohort, high-risk group patients may benefit from erlotinib. Laptinib and Acetalax have primarily been studied for their potential roles in HER2-positive and triple-negative breast cancers, respectively [75–76]. Their effects on pancreatic cancer require further experimental and theoretical support.

Based on our study cohort's enrichment analysis results, the high-risk group is primarily enriched in MYC target-related pathways, DNA repair, G2/M checkpoint-related pathways, mTORC1 signaling, hypoxia, apoptosis-related pathways, the p53 pathway, Notch signaling pathway, and TGF-β signaling-related Hallmark pathways.

MYC is a crucial transcription factor closely associated with cell growth and proliferation. Multiple biological processes regulated by c-MYC enhance ribosome generation and function [58,77]. mTORC1 is a key regulator of cellular metabolism, protein synthesis, and cell proliferation. It enhances metabolic functions by regulating glycolysis, lipid synthesis, and ribosome generation, and supports rapid tumor cell proliferation through protein synthesis [78]. The Notch signaling pathway is believed to regulate tumor development and the tumor microenvironment [79]. Other pathways are closely linked to processes related to cell proliferation and cancer progression. GSEA enrichment analysis also revealed a strong association with cancer cell proliferation pathways related to cell cycle and mitosis. GO-KEGG analysis showed significant enrichment in neuroactive ligand-receptor interactions, transmembrane receptors, ion channels, neurotransmitter-regulated processes, and pathways, indicating a complex interplay between high metabolic activity in tumor cells and neural signaling networks and regulation.

In the cytokine network of pancreatic cancer, immature CD4⁺ T cells can be activated in various cytokine environments [80]. Studies have shown that in pancreatic ductal adenocarcinoma, tumor-associated macrophages (TAMs) rely on IL-10 through the NLRP3 signaling pathway to induce CD4⁺ T cell differentiation into Th2, Th17, and regulatory T cells (Tregs), while inhibiting Th1 polarization and the activation of CD8⁺ cytotoxic T cells [81]. Additionally, cytokines such as TNF-α and IL-1β released by tumor cells can induce cancer-associated fibroblasts (CAFs) to secrete thymic stromal lymphopoietin (TSLP), activating resident dendritic cells (DCs). These DCs release Th2 chemokines (CCL2, CCL17), recruiting Th2 cells to the tumor region. Subsequently, Th2 cells secrete cytokines that promote M2 polarization, mediating immune evasion [82].

These findings are consistent with our study, where high-risk patients in our model cohort exhibited higher infiltration levels of CD4⁺ T cells and Type II helper T cells. In the low-risk group, increased infiltration levels of activated B cells, eosinophils, monocytes, and macrophages are associated with enhanced immune responses and surveillance, as well as pro-inflammatory functions. The M1 subtype of macrophages possesses antitumor activities [83], suggesting that the low-risk group may predominantly harbor M1 macrophages. However, our study also found a positive correlation between M1 macrophages and risk scores, although insufficient to indicate significantly higher infiltration levels in the high-risk group. This slight increase may be related to enhanced metabolism leading to compensatory antitumor activity, warranting further investigation. These results provide a basis for further exploring the potential roles of immune cells in tumor prognosis and therapy.

Tumor immunotherapy, particularly immune checkpoint blockade (ICB) and cell therapy, remains a persistent area of interest in cancer treatment. However, its clinical efficacy is often limited by the immunosuppressive tumor microenvironment. In the TIDE analysis, we found that the differences in responses to ICB treatment between high-risk and low-risk groups did not reach statistical significance. This may reflect the complex immune microenvironment of pancreatic cancer, making it challenging for a single risk score to comprehensively predict responses to ICB treatment. Additionally, limitations of bioinformatics analyses, such as insufficient sample sizes, data heterogeneity, and lack of experimental validation, may also impact the significance of the results. Therefore, future studies should integrate larger-scale clinical data and experimental validations to deeply explore the potential associations between ribosome biogenesis risk scores and ICB treatment responses. Furthermore, our study revealed that multiple immune checkpoint genes, including CD44, LGALS9, CD160, IDO2, and TNFSF9, were significantly more highly expressed in high-risk patients. This suggests that immune checkpoint inhibitors (ICIs) targeting these molecules may offer potential therapeutic benefits. However, ICB treatment prediction indicated that risk scores did not significantly differ in their association with ICB treatment responses, possibly due to the inherently "cold tumor" characteristics of pancreatic cancer, which render it less responsive to immunotherapy [84]. Immunotherapy for pancreatic malignancies still requires further exploration.

In our analysis of somatic mutations, we found that TP53 and KRAS had significantly higher mutation frequencies in the high-risk group compared to the low-risk group. Both subgroups exhibited high mutation rates of CDKN2A and SMAD4, with slightly higher rates in the high-risk group. Mutations in KRAS and TP53 have been extensively studied and consistently shown to have profound impacts on the survival of pancreatic cancer patients. KRAS is an oncogene whose mutations lead to the continuous activation of the RAS/RAF/MEK/ERK signaling pathway, promoting cell proliferation, survival, and migration [85]. TP53 is a tumor suppressor gene whose mutations typically result in cell cycle dysregulation and reduced DNA damage repair capabilities, thereby facilitating tumor development [86]. Mutations in CDKN2A often lead to aberrant cell cycle regulation, while deletions or mutations in SMAD4 are closely associated with metastatic pancreatic cancer [87]. TTN, a structural protein in striated muscle, has the longest exon length in the genome [88]. The predictive significance of TTN mutations may be primarily driven by statistical mechanisms, with limited support in existing literature, and its specific functional mechanisms in certain cancers remain unclear [89]. Future experimental studies are needed to validate how TTN mutations affect the structural stability of pancreatic cancer cells and the specific functions of signaling pathways, thereby clarifying their clinical significance. These findings underscore the critical role of key gene mutations in the progression of pancreatic cancer and suggest that therapeutic strategies targeting these gene mutations may help improve patient prognosis and develop personalized treatment plans.

Although some miRNAs have been identified to play roles in pancreatic cancer cell proliferation, invasion, and metastasis, their study remains limited. In our study cohort, we identified that hsa-miR-203a-3p, hsa-miR-196b-5p, and hsa-miR-205-5p were highly expressed in the high-risk group based on the ribosome biogenesis risk score model, whereas these miRNAs were not highly expressed in the low-risk group.

hsa-miR-203a-3p plays a critical role in various cancers, typically acting as a tumor-suppressive miRNA. In nasopharyngeal carcinoma, miR-203a-3p inhibits tumor growth and metastasis by targeting LASP1 [91]. Its high expression in breast cancer and non-small cell lung cancer is also associated with better prognosis [92,93]. However, our study found that miR-203a-3p is highly expressed in the high-risk group of pancreatic cancer, suggesting it may have dual functions in different tumor types. Existing literature indicates that miR-203a-3p acts as a tumor suppressor in some cancers, while in others, it may promote tumor progression by enhancing epithelial-mesenchymal transition (EMT) and targeting PDE4D to promote colorectal cancer cell proliferation, colony formation, and migration and invasion [94,95]. This dual functionality may be related to its regulation of different target genes and signaling pathways in various biological environments. Therefore, future studies should further investigate the specific mechanisms of miR-203a-3p in pancreatic cancer to elucidate its potential therapeutic value.

miR-196b-5p in pancreatic cancer can regulate the expression of the ING5 gene, inhibiting apoptosis and promoting cancer cell proliferation [96]. It also enhances the invasiveness and anti-apoptotic capabilities of cancer cells in lung small cell carcinoma and colorectal cancer [97,98].

hsa-miR-205-5p exhibits dual roles in different tumors, acting both as a tumor suppressor and a tumor promoter. Its specific function depends on the type of cancer and the tissue microenvironment. Currently, there are few studies confirming miR-205 as a biomarker for pancreatic cancer. Some studies suggest it has ambiguous and pro-tumorigenic roles [99].

Additionally, we analyzed the single-cell sequencing data of the nine genes included in our model and observed differential distribution of cells expressing these molecules in pancreatic cancer. Notably, these cells were primarily distributed in ductal epithelial cells, acinar cells, and a subset of cancer stem cells. Combined with the impact of the risk model on immune infiltration, we speculate that the "cold tumor" characteristics of pancreatic cancer cells may contribute to the formation of an immunosuppressive microenvironment, ultimately resulting in an "immune desert" phenotype. Moreover, our cell communication analysis revealed intricate interactions among cell populations. Ductal epithelial cells associated with hub genes may have complex connections with the MIF signaling pathway, warranting further investigation.

Despite identifying ribosome biogenesis-related risk score groups and prognostic genes in pancreatic ductal adenocarcinoma (PAAD) for the first time, our study has several limitations. First, the number of cohorts and sequencing data used was limited. Second, clinical and pathological feature information of patients was insufficient, necessitating more practical and valuable factors to predict 1- to 5-year survival rates. Lastly, there was a lack of basic experiments to validate the expression of these prognostic genes in pancreatic cancer cell lines, which requires further investigation.

In summary, we identified key prognostic genes related to ribosome biogenesis and constructed a prognostic risk model for PAAD. The results revealed nine prognostic genes that are highly associated with tumorigenesis and immune cell infiltration. This prognostic risk model demonstrated robust and effective performance in predicting overall survival outcomes for PAAD patients. Overall, our findings hold significant implications for in-depth studies of the molecular pathways and mechanisms underlying PAAD, as well as for the development of targeted therapeutic and prognostic strategies for pancreatic cancer.

Limitations

Although this study integrated multiple public databases and employed a series of bioinformatics and machine learning analyses, there are still several limitations. First, all the data used were derived from publicly available databases, potentially leading to issues with data completeness and heterogeneity. Secondly, this study primarily relies on bioinformatics analyses, and further in vitro and in vivo experimental validation of the molecular biological mechanisms of these prognostic genes in pancreatic cancer cell lines is still needed. Finally, given the significant heterogeneity of pancreatic cancer, further validation in multiple centers with diverse patient populations is required to assess the generalizability and clinical applicability of these findings.

Acknowledgements

We would like to thank TCGA ,ICGC and GEO databases for the availability of the data

Author Contribution

Y.S. designed the study concept, performed data analyses, wrote the main manuscript text, and prepared Figures 1–12. Y.L. wrote the main manuscript text and performed data analyses. A.Z. and T.H. verified the data analysis results and prepared Figures 13–17. M.L. provided conceptual guidance and conceived and supervised the study. All authors reviewed the manuscript.

Funding

None.

Data Availability

The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Electronic Supplementary Material

Below is the link to the electronic supplementary material

Supplementary Material 1

Supplementary Material 2

References

1. Klein AP. Pancreatic cancer epidemiology: understanding the role of lifestyle and inherited risk factors. Nat Rev Gastroenterol Hepatol. 2021;18(7):493–502. doi:10.1038/s41575-021-00457-x

2. Mukund A, Afridi MA, Karolak A, Park MA, Permuth JB, Rasool G. Pancreatic Ductal Adenocarcinoma (PDAC): A Review of Recent Advancements Enabled by Artificial Intelligence. Cancers (Basel). 2024;16(12):2240. Published 2024 Jun 17. doi:10.3390/cancers16122240

3. Principe DR, Underwood PW, Korc M, Trevino JG, Munshi HG, Rana A. The Current Treatment Paradigm for Pancreatic Ductal Adenocarcinoma and Barriers to Therapeutic Efficacy. Front Oncol. 2021;11:688377. Published 2021 Jul 15. doi:10.3389/fonc.2021.688377

4. Miller PN, Romero-Hernandez F, Calthorpe L, et al. Long-Duration Neoadjuvant Therapy with FOLFIRINOX Yields Favorable Outcomes for Patients Who Undergo Surgery for Pancreatic Cancer. Ann Surg Oncol. 2024;31(9):6147–6156. doi:10.1245/s10434-024-15579-0

5. Zeng S, Pöttler M, Lan B, Grützmann R, Pilarsky C, Yang H. Chemoresistance in Pancreatic Cancer. Int J Mol Sci. 2019;20(18):4504. Published 2019 Sep 11. doi:10.3390/ijms20184504

6. Melnikov S, Manakongtreecheep K, Söll D. Revising the Structural Diversity of Ribosomal Proteins Across the Three Domains of Life. Mol Biol Evol. 2018;35(7):1588–1598. doi:10.1093/molbev/msy021

7. Kang J, Brajanovski N, Chan KT, Xuan J, Pearson RB, Sanij E. Ribosomal proteins and human diseases: molecular mechanisms and targeted therapy. Signal Transduct Target Ther. 2021;6(1):323. Published 2021 Aug 30. doi:10.1038/s41392-021-00728-8

8. Narla A, Ebert BL. Ribosomopathies: human disorders of ribosome dysfunction. Blood. 2010;115(16):3196–3205. doi:10.1182/blood-2009-10-178129

9. Steinbusch MMF, Caron MMJ, Surtel DAM, et al. Expression of RMRP RNA is regulated in chondrocyte hypertrophy and determines chondrogenic differentiation. Sci Rep. 2017;7(1):6440. Published 2017 Jul 25. doi:10.1038/s41598-017-06809-5

10. Bywater MJ, Poortinga G, Sanij E, et al. Inhibition of RNA polymerase I as a therapeutic strategy to promote cancer-specific activation of p53. Cancer Cell. 2012;22(1):51–65. doi:10.1016/j.ccr.2012.05.019

11. Pylayeva-Gupta Y, Grabocka E, Bar-Sagi D. RAS oncogenes: weaving a tumorigenic web. Nat Rev Cancer. 2011;11(11):761–774. Published 2011 Oct 13. doi:10.1038/nrc3106

12. Azman MS, Alard EL, Dodel M, et al. An ERK1/2-driven RNA-binding switch in nucleolin drives ribosome biogenesis and pancreatic tumorigenesis downstream of RAS oncogene. EMBO J. 2023;42(11):e110902. doi:10.15252/embj.2022110902

13. Zhang Z, Yu H, Yao W, et al. RRP9 promotes gemcitabine resistance in pancreatic cancer via activating AKT signaling pathway. Cell Commun Signal. 2022;20(1):188. Published 2022 Nov 24. doi:10.1186/s12964-022-00974-5

14. Rowell MC, Deschênes-Simard X, Lopes-Paciencia S, et al. Targeting ribosome biogenesis reinforces ERK-dependent senescence in pancreatic cancer. Cell Cycle. 2023;22(19):2172–2193. doi:10.1080/15384101.2023.2278945

15. Stoop TF, Theijse RT, Seelen LWF, et al. Preoperative chemotherapy, radiotherapy and surgical decision-making in patients with borderline resectable and locally advanced pancreatic cancer. Nat Rev Gastroenterol Hepatol. 2024;21(2):101–124. doi:10.1038/s41575-023-00856-2

16. Norris K, Hopes T, Aspden JL. Ribosome heterogeneity and specialization in development. Wiley Interdiscip Rev RNA. 2021;12(4):e1644. doi:10.1002/wrna.1644

17. Doudna JA, Rath VL. Structure and function of the eukaryotic ribosome: the next frontier. Cell. 2002;109(2):153–156. doi:10.1016/s0092-8604(02)00725-0

18. Heiss FB, Daiß JL, Becker P, Engel C. Conserved strategies of RNA polymerase I hibernation and activation. Nat Commun. 2021;12(1):758. Published 2021 Feb 3. doi:10.1038/s41460-021-21031-8

19. Figueiredo VC, McCarthy JJ. Targeting cancer via ribosome biogenesis: the cachexia perspective. Cell Mol Life Sci. 2021;78(15):5775–5787. doi:10.1007/s00018-021-03888-6

20. Wang X, Yu H, Sun W, et al. The long non-coding RNA CYTOR drives colorectal cancer progression by interacting with NCL and Sam68. Mol Cancer. 2018;17(1):110. Published 2018 Jul 31. doi:10.1186/s12943-018-0860-7

21. Khurts S, Masutomi K, Delgermaa L, et al. Nucleolin interacts with telomerase. J Biol Chem. 2004;279(49):51508–51515. doi:10.1074/jbc.M407643200

22. Pellegrino S, Terrosu S, Yusupova G, Yusupov M. Inhibition of the Eukaryotic 80S Ribosome as a Potential Anticancer Therapy: A Structural Perspective. Cancers (Basel). 2021;13(17):4392. Published 2021 Aug 31. doi:10.3390/cancers13174392

23. Elhamamsy AR, Metge BJ, Alsheikh HA, Shevde LA, Samant RS. Ribosome Biogenesis: A Central Player in Cancer Metastasis and Therapeutic Resistance. Cancer Res. 2022;82(13):2344–2353. doi:10.1158/0008-5472.CAN-21-4087

24. Antal CE, Oh TG, Aigner S, et al. A super-enhancer-regulated RNA-binding protein cascade drives pancreatic cancer. Nat Commun. 2023;14(1):5195. Published 2023 Sep 6. doi:10.1038/s41460-023-40798-6

25. Carlos Perez, Mudassier Ahmad, Andrew Massey, Asif Shahriar, Emmanuel Anning, Vivek Kashyap, Neeraj Chauhan, Anupam Dhasmana, Manish Tripathi, Subhash Chauhan, Bilal Hafeez; Abstract 3950: Targeting ribosome biogenesis is a novel strategy to suppress the growth of pancreatic cancer. Cancer Res 15 June 2022; 82 (12_Supplement): 3950.

26. Miki T, Smith CL, Long JE, Eva A, Fleming TP. Oncogene ect2 is related to regulators of small GTP-binding proteins [published correction appears in Nature 1993 Aug 19;364(6439):737]. Nature. 1993;362(6419):462–465. doi:10.1038/362462a0

27. Justilien V, Fields AP. Ect2 links the PKCiota-Par6alpha complex to Rac1 activation and cellular transformation. Oncogene. 2009;28(41):3597–3607. doi:10.1038/onc.2009.217

28. Huff LP, Decristo MJ, Trembath D, et al. The Role of Ect2 Nuclear RhoGEF Activity in Ovarian Cancer Cell Transformation. Genes Cancer. 2013;4(11–12):460–475. doi:10.1177/1947601913514851

29. Justilien V, Jameison L, Der CJ, Rossman KL, Fields AP. Oncogenic activity of Ect2 is regulated through protein kinase C iota-mediated phosphorylation. J Biol Chem. 2011;286(10):8149–8157. doi:10.1074/jbc.M110.196113

30. Justilien V, Lewis KC, Murray NR, Fields AP. Oncogenic Ect2 signaling regulates rRNA synthesis in NSCLC. Small GTPases. 2019;10(5):388–394. doi:10.1080/21541248.2017.1335274

31. Cook DR, Kang M, Martin TD, et al. Aberrant Expression and Subcellular Localization of ECT2 Drives Colorectal Cancer Progression and Growth. Cancer Res. 2022;82(1):90–104. doi:10.1158/0008-5472.CAN-20-4218

32. Li C, Peng Z, Wang Y, et al. Epithelial cell transforming 2 is regulated by Yes-associated protein 1 and mediates pancreatic cancer progression and metastasis. Am J Physiol Gastrointest Liver Physiol. 2021;320(3):G380-G395. doi:10.1152/ajpgi.00185.2020 [38] Wu K, Yan M, Liu T, et al. Creatine kinase B suppresses ferroptosis by phosphorylating GPX4 through a moonlighting function. Nat Cell Biol. 2023;25(5):714–725. doi:10.1038/s41556-023-01133-9 [39] Rajeswari MR, Singh D, Jain A, Ray R. Elevated levels of high-mobility-group chromosomal proteins, HMGA1, in murine skin carcinoma. Cancer Lett. 2001;173(1):93–99. doi:10.1016/s0304-3835(01)00688-7 [40] Ozturk N, Singh I, Mehta A, Braun T, Barreto G. HMGA proteins as modulators of chromatin structure during transcriptional activation. Front Cell Dev Biol. 2014;2:5. Published 2014 Mar 6. doi:10.3389/fcell.2014.00005 [41] Cleynen I, Brants JR, Peeters K, et al. HMGA2 regulates transcription of the Imp2 gene via an intronic regulatory element in cooperation with nuclear factor-kappaB. Mol Cancer Res. 2007;5(4):363–372. doi:10.1158/1541-7786.MCR-06-0331 [42] Zhang S, Mo Q, Wang X. Oncological role of HMGA2 (Review). Int J Oncol. 2019;55(4):775–788. doi:10.3892/ijo.2019.4856 [43] Watanabe S, Ueda Y, Akaboshi S, Hino Y, Sekita Y, Nakao M. HMGA2 maintains oncogenic RAS-induced epithelial-mesenchymal transition in human pancreatic cancer cells. Am J Pathol. 2009;174(3):854–868. doi:10.2353/ajpath.2009.080523 [44] Watanabe S, Ueda Y, Akaboshi S, Hino Y, Sekita Y, Nakao M. HMGA2 maintains oncogenic RAS-induced epithelial-mesenchymal transition in human pancreatic cancer cells. Am J Pathol. 2009;174(3):854–868. doi:10.2353/ajpath.2009.080523 [45] Neumayer G, Belzil C, Gruss OJ, Nguyen MD. TPX2: of spindle assembly, DNA damage response, and cancer. Cell Mol Life Sci. 2014;71(16):3027–3047. doi:10.1007/s00018-014-1582-7 [46] Carter SL, Eklund AC, Kohane IS, Harris LN, Szallasi Z. A signature of chromosomal instability inferred from gene expression profiles predicts clinical outcome in multiple human cancers. Nat Genet. 2006;38(9):1043–1048. doi:10.1038/ng1861 [47] Warner SL, Stephens BJ, Nwokenkwo S, et al. Validation of TPX2 as a potential therapeutic target in pancreatic cancer cells. Clin Cancer Res. 2009;15(21):6519–6528. doi:10.1158/1078-0432.CCR-09-0077

48. Wang Z. ErbB Receptors and Cancer. Methods Mol Biol. 2017;1652:3–35. doi:10.1007/978-1-4939-7219-7_1

49. Andrique L, Fauvin D, El Maassarani M, Colasson H, Vannier B, Séité P. ErbB3(80 kDa), a nuclear variant of the ErbB3 receptor, binds to the Cyclin D1 promoter to activate cell proliferation but is negatively controlled by p14ARF. Cell Signal. 2012;24(5):1074–1085. doi:10.1016/j.cellsig.2012.01.002

50. Tagliaferro M, Rosa P, Bellenchi GC, et al. Nucleolar localization of the ErbB3 receptor as a new target in glioblastoma. BMC Mol Cell Biol. 2022;23(1):13. Published 2022 Mar 7. doi:10.1186/s12860-022-00411-y

51. Reif R, Adawy A, Vartak N, et al. Activated ErbB3 Translocates to the Nucleus via Clathrin-independent Endocytosis, Which Is Associated with Proliferating Cells. J Biol Chem. 2016;291(8):3837–3847. doi:10.1074/jbc.M115.686082

52. Liles JS, Arnoletti JP, Kossenkov AV, et al. Targeting ErbB3-mediated stromal-epithelial interactions in pancreatic ductal adenocarcinoma. Br J Cancer. 2011;105(4):523–533. doi:10.1038/bjc.2011.263

53. Lin T, Ren Q, Zuo W, et al. Valproic acid exhibits anti-tumor activity selectively against EGFR/ErbB2/ErbB3-coexpressing pancreatic cancer via induction of ErbB family members-targeting microRNAs. J Exp Clin Cancer Res. 2019;38(1):150. Published 2019 Apr 8. doi:10.1186/s13046-019-1160-9

54. Cao S, Chen Y, Ren Y, Feng Y, Long S. GLUT1 biological function and inhibition: research advances. Future Med Chem. 2021;13(14):1227–1243. doi:10.4155/fmc-2021-0071

55. Zheng H, Long G, Zheng Y, et al. Glycolysis-Related SLC2A1 Is a Potential Pan-Cancer Biomarker for Prognosis and Immunotherapy [published correction appears in Cancers (Basel). 2023 Jan 18;15(3):586. doi: 10.3390/cancers15030586]. Cancers (Basel). 2022;14(21):5344. Published 2022 Oct 29. doi:10.3390/cancers14215344

56. Li Y, Tang S, Shi X, et al. Metabolic classification suggests the GLUT1/ALDOB/G6PD axis as a therapeutic target in chemotherapy-resistant pancreatic cancer. Cell Rep Med. 2023;4(9):101162. doi:10.1016/j.xcrm.2023.101162

57. Cai K, Chen S, Zhu C, et al. FOXD1 facilitates pancreatic cancer cell proliferation, invasion, and metastasis by regulating GLUT1-mediated aerobic glycolysis. Cell Death Dis. 2022;13(9):765. Published 2022 Sep 3. doi:10.1038/s41419-022-05213-w

58. Acehan D, Petzold C, Gumper I, et al. Plakoglobin is required for effective intermediate filament anchorage to desmosomes. J Invest Dermatol. 2008;128(11):2665 − 2605. doi:10.1038/jid.2008.141

59. Yin L, Li Q, Mrdenovic S, et al. KRT13 promotes stemness and drives metastasis in breast cancer through a plakoglobin/c-Myc signaling pathway. Breast Cancer Res. 2022;24(1):7. Published 2022 Jan 25. doi:10.1186/s13058-022-01502-6

60. Takenaka W, Yokoyama Y, Ikehata K, et al. KRT13 is upregulated in pancreatic cancer stem-like cells and associated with radioresistance. J Radiat Res. 2023;64(2):284–293. doi:10.1093/jrr/rrac091

61. Lin B, Zhou X, Lin S, et al. Epigenetic silencing of PRSS3 provides growth and metastasis advantage for human hepatocellular carcinoma. J Mol Med (Berl). 2017;95(11):1237–1249. doi:10.1007/s00109-017-1578-5

62. Hockla A, Miller E, Salameh MA, Copland JA, Radisky DC, Radisky ES. PRSS3/mesotrypsin is a therapeutic target for metastatic prostate cancer. Mol Cancer Res. 2012;10(12):1555–1566. doi:10.1158/1541-7786.MCR-12-0314

63. Jiang G, Cao F, Ren G, et al. PRSS3 promotes tumour growth and metastasis of human pancreatic cancer. Gut. 2010;59(11):1535–1544. doi:10.1136/gut.2009.200105

64. Nyberg P, Ylipalosaari M, Sorsa T, Salo T. Trypsins and their role in carcinoma growth. Exp Cell Res. 2006;312(8):1219–1228. doi:10.1016/j.yexcr.2005.12.024

65. Sessler RJ, Noy N. A ligand-activated nuclear localization signal in cellular retinoic acid binding protein-II. Mol Cell. 2005;18(3):343–353. doi:10.1016/j.molcel.2005.03.026

66. Noy N. Retinoid-binding proteins: mediators of retinoid action. Biochem J. 2000;348 Pt 3(Pt 3):481–495.

67. Noy N. Between death and survival: retinoic acid in regulation of apoptosis. Annu Rev Nutr. 2010;30:201–217. doi:10.1146/annurev.nutr.28.061807.155509

68. Liu J, Tang L, Chu W, Wei L. Cellular Retinoic Acid Binding Protein 2 (CRABP2), Up-regulated by HPV E6/E7, Leads to Aberrant Activation of the Integrin β1/FAK/ERK Signaling Pathway and Aggravates the Malignant Phenotypes of Cervical Cancer. Biochem Genet. 2024;62(4):2686–2701. doi:10.1007/s10528-023-10568-6

69. Maede Y, Shimizu H, Fukushima T, et al. Differential and common DNA repair pathways for topoisomerase I- and II-targeted drugs in a genetic DT40 repair cell screen panel. Mol Cancer Ther. 2014;13(1):214–220. doi:10.1158/1535-7163.MCT-13-0551

70. Alagoz M, Gilbert DC, El-Khamisy S, Chalmers AJ. DNA repair and resistance to topoisomerase I inhibitors: mechanisms, biomarkers and therapeutic targets. Curr Med Chem. 2012;19(23):3874–3885. doi:10.2174/092986012802002590

71. Stoop TF, Theijse RT, Seelen LWF, et al. Preoperative chemotherapy, radiotherapy and surgical decision-making in patients with borderline resectable and locally advanced pancreatic cancer. Nat Rev Gastroenterol Hepatol. 2024;21(2):101–124. doi:10.1038/s41575-023-00856-2

72. Zhao S, Tang Y, Wang R, Najafi M. Mechanisms of cancer cell death induction by paclitaxel: an updated review. Apoptosis. 2022;27(9–10):647–660. doi:10.1007/s10495-022-01750-z

73. Chun JW, Lee SH, Kim JS, et al. Comparison between FOLFIRINOX and gemcitabine plus nab-paclitaxel including sequential treatment for metastatic pancreatic cancer: a propensity score matching approach. BMC Cancer. 2021;21(1):537. Published 2021 May 11. doi:10.1186/s12885-021-08277-7

74. Wang Y, Hu GF, Zhang QQ, et al. Efficacy and safety of gemcitabine plus erlotinib for locally advanced or metastatic pancreatic cancer: a systematic review and meta-analysis. Drug Des Devel Ther. 2016;10:1961–1972. Published 2016 Jun 13. doi:10.2147/DDDT.S105442

75. Suppan C, Balic M. Current Standards and Future Outlooks in Metastatic Her2-Positive Breast Cancer. Breast Care (Basel). 2023;18(1):69–75. doi:10.1159/000528756

76. Mizunuma M, Redon CE, Saha LK, et al. Acetalax (Oxyphenisatin Acetate, NSC 59687) and Bisacodyl Cause Oncosis in Triple-Negative Breast Cancer Cell Lines by Poisoning the Ion Exchange Membrane Protein TRPM4. Cancer Res Commun. 2024;4(8):2101–2111. doi:10.1158/2760-9764.CRC-24-0093

77. Jha RK, Kouzine F, Levens D. MYC function and regulation in physiological perspective. Front Cell Dev Biol. 2023;11:1268275. Published 2023 Oct 24. doi:10.3389/fcell.2023.1268275

78. Yecies JL, Manning BD. mTOR links oncogenic signaling to tumor cell metabolism. J Mol Med (Berl). 2011;89(3):221–228. doi:10.1007/s00109-011-0726-6

79. Meurette O, Mehlen P. Notch Signaling in the Tumor Microenvironment. Cancer Cell. 2018;34(4):536–548. doi:10.1016/j.ccell.2018.07.009

80. Qiu Y, Chen T, Hu R, et al. Next frontier in tumor immunotherapy: macrophage-mediated immune evasion. Biomark Res. 2021;9(1):72. Published 2021 Oct 9. doi:10.1186/s40364-021-00327-3

81. Daley D, Mani VR, Mohan N, et al. NLRP3 signaling drives macrophage-induced adaptive immune suppression in pancreatic carcinoma. J Exp Med. 2017;214(6):1711–1724. doi:10.1084/jem.20161707

82. Halstead SB, Mahalingam S, Marovich MA, Ubol S, Mosser DM. Intrinsic antibody-dependent enhancement of microbial infection in macrophages: disease regulation by immune complexes. Lancet Infect Dis. 2010;10(10):712–722. doi:10.1016/S1473-3099(10)70166-3

83. Liu J, Geng X, Hou J, Wu G. New insights into M1/M2 macrophages: key modulators in cancer progression. Cancer Cell Int. 2021;21(1):389. Published 2021 Jul 21. doi:10.1186/s12935-021-02089-2

84. Barriga FM, Tsanov KM, Ho YJ, et al. MACHETE identifies interferon-encompassing chromosome 9p21.3 deletions as mediators of immune evasion and metastasis. Nat Cancer. 2022;3(11):1360–1385. doi:10.1038/s43018-022-00443-5

85. Mondal K, Posa MK, Shenoy RP, Roychoudhury S. KRAS Mutation Subtypes and Their Association with Other Driver Mutations in Oncogenic Pathways. Cells. 2024;13(14):1221. Published 2024 Jul 19. doi:10.3390/cells13141221

86. Negrini S, Gorgoulis VG, Halazonetis TD. Genomic instability–an evolving hallmark of cancer. Nat Rev Mol Cell Biol. 2010;11(3):220–228. doi:10.1038/nrm2858

87. Hu C, Hart SN, Polley EC, et al. Association Between Inherited Germline Mutations in Cancer Predisposition Genes and Risk of Pancreatic Cancer. JAMA. 2018;319(23):2401–2409. doi:10.1001/jama.2018.6228

88. Ciferri A, Crumbliss AL. The Assembling and Contraction Mechanisms of Striated Muscles. Front Chem. 2018;6:570. Published 2018 Nov 30. doi:10.3389/fchem.2018.00570

89. Jia Q, Wang J, He N, He J, Zhu B. Titin mutation associated with responsiveness to checkpoint blockades in solid tumors. JCI Insight. 2019;4(10):e127901. Published 2019 May 16. doi:10.1172/jci.insight.127901

90. Langfelder P, Horvath S. WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics. 2008;9:559. Published 2008 Dec 29. doi:10.1186/1471-2105-9-559

91. Jiang N, Jiang X, Chen Z, et al. MiR-203a-3p suppresses cell proliferation and metastasis through inhibiting LASP1 in nasopharyngeal carcinoma. J Exp Clin Cancer Res. 2017;36(1):138. Published 2017 Oct 5. doi:10.1186/s13046-017-0604-3

92. Wang L, Wang G, Song J, Yao D, Wang Y, Chen T. A comprehensive analysis of the prognostic characteristics of microRNAs in breast cancer. Front Genet. 2024;15:1293824. Published 2024 Mar 20. doi:10.3389/fgene.2024.1293824

93. Zhang Y, Mi K, Li Z, et al. Identification of Prognostic miRNAs Associated With Immune Cell Tumor Infiltration Predictive of Clinical Outcomes in Patients With Non-Small Cell Lung Cancer. Front Oncol. 2021;11:705869. Published 2021 Jul 1. doi:10.3389/fonc.2021.705869

94. Tarazona JGR, Abdallah EA, Flores BCT, et al. MIR-203A-3P AND MMP-2 PROTEINS ARE HIGHLY EXPRESSED IN CIRCULATING TUMOR CELLS FROM PATIENTS WITH PANCREATIC CARCINOMA. PROTEÍNAS MIR-203A-3P E MMP-2 SÃO ALTAMENTE EXPRESSAS EM CÉLULAS TUMORAIS CIRCULANTES DE PACIENTES COM CARCINOMA PANCREÁTICO. Arq Bras Cir Dig. 2022;34(4):e1628. Published 2022 Jan 31. doi:10.1590/0102-672020210002e1628

95. Chen L, Gao H, Liang J, et al. miR-203a-3p promotes colorectal cancer proliferation and migration by targeting PDE4D. Am J Cancer Res. 2018;8(12):2387–2401. Published 2018 Dec 1.

96. Ma D, Chen S, Wang H, et al. Baicalein Induces Apoptosis of Pancreatic Cancer Cells by Regulating the Expression of miR-139-3p and miR-196b-5p. Front Oncol. 2021;11:653061. Published 2021 Apr 30. doi:10.3389/fonc.2021.653061

97. Xin H, Wang C, Chi Y, Liu Z. MicroRNA-196b-5p promotes malignant progression of colorectal cancer by targeting ING5. Cancer Cell Int. 2020;20:119. Published 2020 Apr 10. doi:10.1186/s12935-020-01200-3

98. Huang X, Xiao S, Zhu X, et al. miR-196b-5p-mediated downregulation of FAS promotes NSCLC progression by activating IL6-STAT3 signaling. Cell Death Dis. 2020;11(9):785. Published 2020 Sep 22. doi:10.1038/s41419-020-02997-7

99. Chauhan N, Dhasmana A, Jaggi M, Chauhan SC, Yallapu MM. miR-205: A Potential Biomedicine for Cancer Therapy. Cells. 2020;9(9):1957. Published 2020 Aug 25. doi:10.3390/cells9091957

Yes