Integrative prognostic modeling of ovarian cancer: incorporating genetic, clinical, and immunological markers.

Initially, endometriosis datasets were downloaded from the GEO database, and RNA sequencing data along with associated clinical information for 379 ovarian cancer samples were collected from The Cancer Genome Atlas (TCGA) at https://portal.gdc.cancer.gov/ . This information included treatment details, somatic mutations, copy number variations, and data on 88 normal ovarian samples from the GTEx database. The data were randomly divided into training and testing groups in a 1:1 ratio. Statistical analyses were conducted based on clinical variables such as tumor staging, grading, and patient age. All information and clinical matrices involved in this study were downloaded from public databases. No ethical committee approval or written informed consent from patients was required. Differential analysis of gene matrices for endometriosis and ovarian cancer was conducted using the limma package, and heatmaps were generated using the pheatmap package. Single-factor Cox regression analysis was performed on endometriosis-related differential genes using the survival package. Consensus clustering analysis was then used to classify ovarian cancer patients into two subtypes based on prognostically relevant genes. The ConsensusClusterPlus package was employed for clustering analysis. The limma package was used to analyze genes between the two subtypes, with selection criteria of |logFC|> 1 and an adjusted p-value < 0.05. The clusterProfiler package was utilized for Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis of the differential genes. Enrichment analysis of genomic variations was conducted using the GSVA package to study biological processes. Based on TCGA's somatic mutation data, the maftools R package was used to analyze gene mutations in ovarian cancer patients. Immune cell infiltration analysis was conducted using ssGSEA, which included 28 types of immune cells. The estimate package was utilized to calculate immune scores, including immune and stromal scores, estimate scores, and tumor purity scores. Limma was used to compare the expression of immune checkpoint inhibitors between subtypes, and differences in subtype were displayed using box plots generated by ggplot. Initially, candidate genes were narrowed down using the glmnet package, applying the least absolute shrinkage and selection operator (LASSO) Cox regression model, based on the results from univariate Cox regression analysis. Multivariable Cox analysis was then conducted to establish characteristics using the clusterSur package. All samples were divided into low-risk and high-risk groups based on the median risk score. Survival curves were used to compare the survival outcomes between these two groups. The R packages riskplot and ROC were utilized for risk factor analysis and risk curve analysis, respectively. Additionally, survival rates for the high-risk and low-risk groups were assessed. Time-dependent ROC curves for 1-year, 3-year, and 5-year survival rates were also employed to evaluate the capability of this feature to predict the prognosis of ovarian cancer patients in high-risk and low-risk groups. Univariate and multivariate analyses were performed to determine whether the risk score could serve as an independent prognostic factor, distinct from other clinical and pathological factors. Additionally, nomograms were constructed using the rms package to predict 1-year, 3-year, and 5-year survival rates for ovarian cancer patients. Finally, decision curve analysis (DCA), calibration curves, and concordance index (C-index) were plotted to assess the accuracy of the nomograms. Using somatic mutation data from TCGA, gene mutations in ovarian cancer patients were reanalyzed with the maftools R package, which also calculated the tumor mutational burden (TMB) for each patient. Differences in TMB between high-risk and low-risk groups were compared. Furthermore, survival analysis based on TMB scores was conducted for patients, providing insights into the relationship between TMB and patient outcomes. The immunocor R package was utilized to quantify immune cell infiltration in both high-risk and low-risk groups. To enhance the resolution of immune profiling, ssGSEA was expanded to evaluate specific immune cell subtypes, including CD8 + T cells, regulatory T cells, and macrophage subtypes (M1/M2). Additionally, we assessed immune functionality through the expression of key immune checkpoint molecules, such as PD-1, CTLA-4, and their ligands, to characterize the immunosuppressive environment. Tumor mutational burden (TMB) and predicted neoantigen load were integrated into the analysis to explore their relationship with immune responses and prognostic significance. Spearman's correlation analysis was used to evaluate the differential expression of immune checkpoint molecules and HLA-related genes between the high-risk and low-risk groups. The Genomics of Drug Sensitivity in Cancer database was used to estimate the chemotherapy response of ovarian cancer patients to chemotherapeutic drugs. The oncoPredict R package was utilized to determine the half-maximal inhibitory concentration (IC50) of anticancer drugs for the high-risk and low-risk groups.

A differential analysis of endometriosis was initially conducted, selecting genes with a log fold change (logFC) greater than 1 and an adjusted p-value of less than 0.05, resulting in 1,263 differential genes. A heatmap of the top 20 differential genes is shown (Fig.  1 A). The expression matrix of these 1,263 genes was then extracted from ovarian cancer datasets, and differential analysis was performed under the same criteria, identifying 580 differential genes. A heatmap displays the ovarian cancer differential genes related to endometriosis (Fig.  1 B). Subsequent enrichment analysis on these 580 genes revealed their involvement in biological functions like organelle fission and collagen-containing extracellular matrix through GO analysis (Fig.  1 C), and pathways such as the p53 signaling pathway and PI3K-Akt signaling pathway through KEGG analysis (Fig.  1 D), indicating a role in malignant tumor development. Fig. 1 Differential Gene Expression and Pathway Analysis in Endometriosis and Ovarian Cancer. A Heatmap displaying the top 20 differentially expressed genes in endometriosis datasets, filtered by a log fold change (logFC) greater than 1 and an adjusted p-value less than 0.05. B Heatmap of 580 differentially expressed genes in ovarian cancer related to endometriosis, identified under the same filtering criteria as Figure A . C Gene Ontology (GO) enrichment analysis of the 580 differentially expressed genes, highlighting significant involvement in biological functions such as organelle fission and collagen-containing extracellular matrix. D Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis showing that these genes are predominantly enriched in pathways such as the p53 signaling pathway and PI3K-Akt signaling pathway, suggesting their role in malignancy associated with endometriosis Differential Gene Expression and Pathway Analysis in Endometriosis and Ovarian Cancer. A Heatmap displaying the top 20 differentially expressed genes in endometriosis datasets, filtered by a log fold change (logFC) greater than 1 and an adjusted p-value less than 0.05. B Heatmap of 580 differentially expressed genes in ovarian cancer related to endometriosis, identified under the same filtering criteria as Figure A . C Gene Ontology (GO) enrichment analysis of the 580 differentially expressed genes, highlighting significant involvement in biological functions such as organelle fission and collagen-containing extracellular matrix. D Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis showing that these genes are predominantly enriched in pathways such as the p53 signaling pathway and PI3K-Akt signaling pathway, suggesting their role in malignancy associated with endometriosis To identify genes associated with ovarian cancer prognosis, univariate Cox regression analysis was applied to the 580 differential genes, finding 54 prognostically relevant genes (Fig.  2 A). Consensus clustering of these 54 endometriosis-related prognostic genes at k = 2 effectively differentiated ovarian cancer samples into groups (Figs.  2 B–D). Differential analysis of these genes between the two subtypes indicated substantial variations, as shown in the heatmap (Fig.  2 E), leading to the identification of 101 differential genes (Fig.  2 F). Further analyses, including GO and KEGG enrichment, were performed to explore the biological functions and pathways of these subtype-specific differential genes, revealing significant involvement in extracellular matrix organization and the TGF-beta signaling pathway (Figs.  2 G, H ). Fig. 2 Prognostic Gene Analysis and Subtype Clustering in Ovarian Cancer. A Forest plot showing the 54 genes associated with the prognosis of ovarian cancer patients, identified through univariate Cox regression analysis. B – D Results of consensus clustering analysis indicating optimal patient stratification into two subtypes at k = 2. 2E: Heatmap illustrating significant gene expression differences between the two subtypes for the 54 prognostic genes. 2 F : Identification of 101 differential genes between the two subtypes with highlighted biological functions in GO and KEGG pathway analyses shown in Figures G and H . G O analysis results showing that differential genes are mainly involved in biological functions such as extracellular matrix organization and collagen-containing extracellular matrix. 2H: KEGG pathway analysis revealing enrichment in pathways like TGF-beta signaling, indicating their involvement in subtype-specific cancer processes Prognostic Gene Analysis and Subtype Clustering in Ovarian Cancer. A Forest plot showing the 54 genes associated with the prognosis of ovarian cancer patients, identified through univariate Cox regression analysis. B – D Results of consensus clustering analysis indicating optimal patient stratification into two subtypes at k = 2. 2E: Heatmap illustrating significant gene expression differences between the two subtypes for the 54 prognostic genes. 2 F : Identification of 101 differential genes between the two subtypes with highlighted biological functions in GO and KEGG pathway analyses shown in Figures G and H . G O analysis results showing that differential genes are mainly involved in biological functions such as extracellular matrix organization and collagen-containing extracellular matrix. 2H: KEGG pathway analysis revealing enrichment in pathways like TGF-beta signaling, indicating their involvement in subtype-specific cancer processes Comparative analysis was conducted between two subtypes with respect to survival and immunological features. Survival curve analysis indicated that patients with the ES low subtype had a better prognosis (Fig.  3 A). GSVA analysis identified significant involvement in pathways such as GAP_JUNCTION among other KEGG pathways (Fig.  3 B). ssGSEA analysis of immune cell infiltration showed higher infiltration of various immune cells in the ES high subtype (Fig.  3 C). The immune scoring analysis revealed that the ES high subtype had higher immune-related scores, whereas the ES low subtype had higher tumor purity scores (Figs.  3 D–G). Analysis of immune checkpoint inhibitors indicated higher expression of multiple inhibitors in the ES high subtype (Fig.  3 H). Fig. 3 Survival, Immune Analysis, and Immune Checkpoint Expression in Ovarian Cancer Subtypes. A Kaplan–Meier survival curves showing better prognosis for patients with the ES low subtype compared to the ES high subtype. B Gene Set Variation Analysis (GSVA) results identifying involvement in key KEGG pathways like GAP_JUNCTION, which are differentially active between the subtypes. C Single-sample Gene Set Enrichment Analysis (ssGSEA) of immune cell infiltration indicating higher infiltration levels of various immune cells in the ES high subtype. D – G Immune scoring analysis results displaying higher immune-related scores in the ES high subtype and higher tumor purity scores in the ES low subtype. 3H: Expression levels of various immune checkpoint inhibitors are higher in the ES high subtype, suggesting a potential for targeted immunotherapy in this group Survival, Immune Analysis, and Immune Checkpoint Expression in Ovarian Cancer Subtypes. A Kaplan–Meier survival curves showing better prognosis for patients with the ES low subtype compared to the ES high subtype. B Gene Set Variation Analysis (GSVA) results identifying involvement in key KEGG pathways like GAP_JUNCTION, which are differentially active between the subtypes. C Single-sample Gene Set Enrichment Analysis (ssGSEA) of immune cell infiltration indicating higher infiltration levels of various immune cells in the ES high subtype. D – G Immune scoring analysis results displaying higher immune-related scores in the ES high subtype and higher tumor purity scores in the ES low subtype. 3H: Expression levels of various immune checkpoint inhibitors are higher in the ES high subtype, suggesting a potential for targeted immunotherapy in this group To construct a prognostic model, lasso regression analysis and multivariate Cox regression analysis were performed based on the results of univariate Cox regression analysis. The lasso regression analysis results, shown in Fig.  4 A, B, included 17 genes for further analysis. The multivariate Cox regression analysis identified 10 genes associated with the prognosis of ovarian cancer patients. A risk score was established based on the regression coefficients of these 10 genes. Patients in the TCGA ovarian cancer dataset were randomly divided into training, validation, and TCGA cohorts in a 1:1 ratio. Risk factor analysis showed that patients with higher risk scores had higher mortality rates, consistently across all datasets; heatmaps demonstrated significant differences between high and low-risk groups in model genes (Fig.  4 C–E). Fig. 4 LASSO Regression and Risk Score Analysis for Ovarian Cancer. A LASSO coefficient profiles of the variables against the log(lambda) sequence, indicating the selection of the most optimal predictors for the prognostic model. B Cross-validation for tuning parameter selection in the LASSO model. The red dots represent partial likelihood deviance with varying log(lambda) values, showing minimum criteria and 1-standard error rule. C – E Heatmaps displaying gene expression profiles in ovarian cancer patients, segmented by risk scores into high and low-risk groups. These heatmaps illustrate significant expression differences across the model genes between the two risk groups, reinforcing the discriminative power of the risk score LASSO Regression and Risk Score Analysis for Ovarian Cancer. A LASSO coefficient profiles of the variables against the log(lambda) sequence, indicating the selection of the most optimal predictors for the prognostic model. B Cross-validation for tuning parameter selection in the LASSO model. The red dots represent partial likelihood deviance with varying log(lambda) values, showing minimum criteria and 1-standard error rule. C – E Heatmaps displaying gene expression profiles in ovarian cancer patients, segmented by risk scores into high and low-risk groups. These heatmaps illustrate significant expression differences across the model genes between the two risk groups, reinforcing the discriminative power of the risk score Prognostic analysis across three cohorts showed that patients with a higher risk score had worse outcomes compared to those with a low risk (Figs.  5 A, C, E). Finally, to validate the prognostic efficacy of the model, ROC curve analysis was conducted. The analysis revealed AUC values of 0.642, 0.611, and 0.627 for 1-year; 0.751, 0.602, and 0.677 for 3-years; and 0.790, 0.622, and 0.699 for 5-years, across the three datasets respectively (Figs.  5 B, D, F). Fig. 5 Survival Analysis and ROC Curve Evaluation for Prognostic Model. A , C , E Kaplan–Meier survival curves for patients stratified into high and low-risk groups based on the prognostic model, showing significantly different outcomes with p-values < 0.001. B , D , F ROC curves assessing the predictive accuracy of the prognostic model at 1-year, 3-year, and 5-year intervals for three different datasets. AUC metrics indicate robust model performance in predicting patient outcomes over time Survival Analysis and ROC Curve Evaluation for Prognostic Model. A , C , E Kaplan–Meier survival curves for patients stratified into high and low-risk groups based on the prognostic model, showing significantly different outcomes with p-values < 0.001. B , D , F ROC curves assessing the predictive accuracy of the prognostic model at 1-year, 3-year, and 5-year intervals for three different datasets. AUC metrics indicate robust model performance in predicting patient outcomes over time To evaluate the independent predictive ability of the risk score, univariate and multivariate Cox regression analyses were performed in conjunction with clinical characteristics. The results confirmed that the risk score independently predicts outcomes in both univariate and multivariate analyses (Figs.  6 A, B ). A nomogram was then constructed to predict survival rates, revealing a 1-year survival rate of 95.1% (Fig.  6 C). Calibration curves, concordance index (C-index), and decision curve analysis (DCA) indicated good predictive performance of the nomogram (Figs.  6 D–F). Comparison of risk scores between different ES subtypes showed that the ES high subtype had higher risk scores (Fig.  6 G). Sankey diagrams illustrated that the ES high subtype is associated with higher risk scores and worse clinical staging (Fig.  6 H). Fig. 6 Validation of Risk Score's Independent Predictive Power and Nomogram Construction. A , B Forest plots of univariate and multivariate Cox regression analyses demonstrating the independent prognostic significance of the risk score alongside other clinical features such as age, grade, and stage. C Nomogram for predicting the 1-year survival rate based on risk score and clinical parameters, with the observed 1-year survival rate at 95.1%. D - F Calibration curve, concordance index (C-index), and decision curve analysis (DCA) confirming the predictive accuracy and clinical usefulness of the nomogram. G Distribution of risk scores between different ES subtypes, highlighting significantly higher risk scores in the ES high subtype. H Sankey diagram showing the flow of patients from ES subtype classification through risk score to clinical staging, emphasizing worse clinical stages associated with higher risk scores Validation of Risk Score's Independent Predictive Power and Nomogram Construction. A , B Forest plots of univariate and multivariate Cox regression analyses demonstrating the independent prognostic significance of the risk score alongside other clinical features such as age, grade, and stage. C Nomogram for predicting the 1-year survival rate based on risk score and clinical parameters, with the observed 1-year survival rate at 95.1%. D - F Calibration curve, concordance index (C-index), and decision curve analysis (DCA) confirming the predictive accuracy and clinical usefulness of the nomogram. G Distribution of risk scores between different ES subtypes, highlighting significantly higher risk scores in the ES high subtype. H Sankey diagram showing the flow of patients from ES subtype classification through risk score to clinical staging, emphasizing worse clinical stages associated with higher risk scores Analysis of prognostic differences between risk groups found that age, grade, and stage were closely related to the risk score. It was observed that patients with low risk had better prognoses compared to those in the high-risk group, after grouping according to clinical characteristics (Fig.  7 ). Fig. 7 Prognostic Differences Between Risk Groups Stratified by Clinical Features. A Kaplan–Meier survival curves illustrating the survival outcomes of ovarian cancer patients stratified by age. Patients aged ≥ 65 in the low-risk group show significantly better survival compared to those in the high-risk group, highlighting the prognostic influence of the risk score across different age groups. B Kaplan–Meier survival curves for patients stratified by tumor grade. The survival differences between the high and low-risk groups are evident within each grade category (G1&G2 vs. G3&G4), with low-risk patients exhibiting notably better survival outcomes. C Survival analysis of patients stratified by tumor stage (Stage I&II vs. Stage III&IV). The plots show a clear separation in survival outcomes, with patients in early stages (I&II) in the low-risk group having a much better prognosis than those in the high-risk group, demonstrating the utility of the risk score in predicting survival across different clinical stages. Figure 7 Comparative Prognosis Across Different Risk Groups by Clinical Characteristics Prognostic Differences Between Risk Groups Stratified by Clinical Features. A Kaplan–Meier survival curves illustrating the survival outcomes of ovarian cancer patients stratified by age. Patients aged ≥ 65 in the low-risk group show significantly better survival compared to those in the high-risk group, highlighting the prognostic influence of the risk score across different age groups. B Kaplan–Meier survival curves for patients stratified by tumor grade. The survival differences between the high and low-risk groups are evident within each grade category (G1&G2 vs. G3&G4), with low-risk patients exhibiting notably better survival outcomes. C Survival analysis of patients stratified by tumor stage (Stage I&II vs. Stage III&IV). The plots show a clear separation in survival outcomes, with patients in early stages (I&II) in the low-risk group having a much better prognosis than those in the high-risk group, demonstrating the utility of the risk score in predicting survival across different clinical stages. Figure 7 Comparative Prognosis Across Different Risk Groups by Clinical Characteristics The impact of immune cell infiltration on tumor progression and prognosis was analyzed in ovarian cancer samples. Heatmap results showed higher infiltration in various immune infiltration algorithms in the high-risk group (Fig.  8 A). Analysis of immune functions between different risk groups revealed higher expression of various immune functions such as CCR in the high-risk group (Fig.  8 B). Further analysis showed that high-risk patients exhibited higher levels of regulatory T cells and M2 macrophages, indicative of an immunosuppressive environment (Fig.  8 C). Conversely, low-risk patients had higher levels of CD8 + T cells and M1 macrophages, correlating with a more active immune state (Fig.  8 C). The integration of TMB and neoantigen load further emphasized the interplay between genetic mutations and immune responses, revealing that high-risk patients with high TMB had worse outcomes, likely due to concurrent immunosuppression (Fig.  8 D–G). Fig. 8 Immune Cell Infiltration and Function Analysis in Ovarian Cancer Subtypes. A Heatmaps illustrating the immune cell infiltration profiles in high and low-risk ovarian cancer patient groups, using various immune infiltration algorithms. High-risk groups show increased infiltration across multiple algorithms. B Analysis of immune functions such as CCR expression, indicating higher expression levels in the high-risk group compared to the low-risk group. C Single-sample Gene Set Enrichment Analysis (ssGSEA) results showing higher levels of various immune cell types in the high-risk group. D – G Immune scoring analysis results demonstrating higher immune-related scores in the high-risk group and higher tumor purity scores in the ES low subtype Immune Cell Infiltration and Function Analysis in Ovarian Cancer Subtypes. A Heatmaps illustrating the immune cell infiltration profiles in high and low-risk ovarian cancer patient groups, using various immune infiltration algorithms. High-risk groups show increased infiltration across multiple algorithms. B Analysis of immune functions such as CCR expression, indicating higher expression levels in the high-risk group compared to the low-risk group. C Single-sample Gene Set Enrichment Analysis (ssGSEA) results showing higher levels of various immune cell types in the high-risk group. D – G Immune scoring analysis results demonstrating higher immune-related scores in the high-risk group and higher tumor purity scores in the ES low subtype Analysis of tumor mutation rates between the two risk groups showed that the low-risk group had a mutation rate of 97.6%, while the high-risk group had a rate of 94.44% (Figs.  9 A, B ). TMB results indicated that the low-risk group had a higher TMB (Fig.  9 C). It was observed that patients with high mutation burden had better prognoses; however, combining risk scores showed that patients with high tumor mutation burden and in the high-risk group had the worst prognosis (Figs.  9 D, E ). Immune checkpoint inhibitor gene analysis found higher expression of common genes including CD276 in the high-risk group compared to the low-risk group (Fig.  9 F). TIDE results indicated lower TIDE scores in the low-risk group (Fig.  9 G). Analysis of immune response to treatment found 66% response in the high-risk group compared to 59% in the low-risk group (Fig.  9 H). There was no significant difference in immune response based on risk score (Fig.  9 I). Fig. 9 Tumor Mutational Burden (TMB) and Response to Immunotherapy. A – B Mutation rate analysis revealing a lower mutation rate in the high-risk group (94.44%) compared to the low-risk group (97.6%). C Analysis showing higher TMB in the low-risk group. D – E Kaplan–Meier survival curves comparing patient outcomes based on high and low TMB within different risk groups, illustrating that high-risk patients with high TMB have the worst prognosis. F Analysis of immune checkpoint inhibitor genes such as CD276, which are expressed higher in the high-risk group. G TIDE scores indicating lower immune evasion (better prognosis) in the low-risk group. H Immunotherapy response rates showing 66% responsiveness in the high-risk group compared to 59% in the low-risk group. I Analysis showing no significant difference in immunotherapy response based on risk score alone Tumor Mutational Burden (TMB) and Response to Immunotherapy. A – B Mutation rate analysis revealing a lower mutation rate in the high-risk group (94.44%) compared to the low-risk group (97.6%). C Analysis showing higher TMB in the low-risk group. D – E Kaplan–Meier survival curves comparing patient outcomes based on high and low TMB within different risk groups, illustrating that high-risk patients with high TMB have the worst prognosis. F Analysis of immune checkpoint inhibitor genes such as CD276, which are expressed higher in the high-risk group. G TIDE scores indicating lower immune evasion (better prognosis) in the low-risk group. H Immunotherapy response rates showing 66% responsiveness in the high-risk group compared to 59% in the low-risk group. I Analysis showing no significant difference in immunotherapy response based on risk score alone Finally, the sensitivity to chemotherapy drugs was analyzed. Results indicated that including platinum-based drugs and Axitinib, the half-maximal inhibitory concentration (ID50) was higher in the high-risk group, suggesting that patients in the low-risk group are more sensitive to chemotherapy drugs (Figs.  10 A–F). Fig. 10 Chemotherapy Drug Sensitivity Analysis. A – F Analysis of chemotherapeutic drug sensitivity, showing higher ID50 values (indicating lower sensitivity) in the high-risk group for drugs including platinum-based therapies and Axitinib, suggesting that low-risk patients are more chemosensitive Chemotherapy Drug Sensitivity Analysis. A – F Analysis of chemotherapeutic drug sensitivity, showing higher ID50 values (indicating lower sensitivity) in the high-risk group for drugs including platinum-based therapies and Axitinib, suggesting that low-risk patients are more chemosensitive

This study aimed to develop an integrative prognostic model combining genetic, clinical, and immunological data to improve outcome prediction for ovarian cancer patients. By analyzing data from The Cancer Genome Atlas (TCGA), we identified key prognostic genes and developed a risk score that effectively stratifies patients into high- and low-risk groups with significant survival differences. Unlike previous models, which primarily focused on genetic or clinical parameters, our model incorporates immune landscape data, including immune cell infiltration and checkpoint expression, to provide a more comprehensive understanding of ovarian cancer progression and treatment response. These findings not only validate the prognostic significance of the tumor microenvironment but also highlight potential therapeutic targets, offering a robust tool for personalized treatment strategies. Several studies have previously attempted to construct prognostic models for ovarian cancer; however, most focused primarily on genetic or clinical parameters independently. For instance, Konecny et al. developed a prognostic model based on genetic alterations in ovarian cancer, which significantly enhanced the prediction of patient outcomes compared to traditional clinical models alone[ 28 ]. While this model was groundbreaking, it did not incorporate the immunological context of the tumor microenvironment, which has been shown to play a critical role in cancer progression and response to treatment. In contrast, our study's inclusion of immune cell infiltration and immune function data represents a significant advancement. The analysis of immune cell presence and activity, as seen through ssGSEA and immune checkpoint expression, provides a deeper insight into the tumor’s interaction with the host immune system. This approach is supported by recent studies suggesting that the immune microenvironment can influence the efficacy of standard treatments and the success of emerging therapies like immunotherapy. By integrating these immunological insights with genetic and clinical data, our model aligns with the shift towards a more holistic view of cancer that encompasses the dynamic interactions within the tumor microenvironment. The risk score developed in this study demonstrated strong predictive power for patient outcomes, as evidenced by survival analyses and validation across multiple cohorts. High-risk patients identified by our model were associated with poor prognosis, a finding consistent with studies that have linked aggressive molecular profiles and immune evasion mechanisms with advanced disease stages and resistance to treatment. To further refine the immune analysis, we evaluated specific immune cell subtypes, such as CD8 + T cells, regulatory T cells, and macrophage polarization (M1/M2), revealing distinct differences in immune cell infiltration between high- and low-risk groups. Additionally, integrating immune checkpoint expression (e.g., PD-1, CTLA-4) and neoantigen load into the analysis highlighted a more suppressive immune environment in high-risk patients, suggesting potential responsiveness to immune checkpoint blockade therapies. These refinements provide deeper insights into the role of the tumor microenvironment and its contribution to prognosis, further enhancing the model’s clinical relevance and potential for guiding personalized treatment strategies. Furthermore, our study expanded on these findings by correlating high-risk scores with increased expression of immune checkpoint inhibitors like PD-L1, suggesting potential responsiveness to checkpoint blockade therapies. This correlation aligns with recent clinical trials that have explored the efficacy of PD-1 and PD-L1 inhibitors in ovarian cancer, emphasizing the importance of predictive biomarkers for selecting suitable candidates for immunotherapy [ 29 , 30 ]. Our study’s emphasis on the differential immune profiles between high and low-risk groups provides a valuable framework for understanding the variability in treatment responses among ovarian cancer patients. For example, the high-risk group exhibited a more suppressive immune environment, which could explain the lower effectiveness of certain therapies in these patients. These findings are in line with research by Zhang et al., which demonstrated that a suppressive immune microenvironment could hinder the efficacy of both chemotherapy and immunotherapy in ovarian cancer [ 31 ]. By analyzing the immune profiles in conjunction with genetic data, our model provides a comprehensive overview that can inform more targeted and effective therapeutic strategies. For instance, patients with high immune cell infiltration but low mutational burden might benefit from therapies that aim to activate the immune response, whereas those with high mutational burden could be better candidates for immunotherapy. The analysis of tumor mutational burden (TMB) in our study highlighted its potential as a prognostic and predictive marker in ovarian cancer. High TMB was associated with better outcomes in low-risk patients, which supports the hypothesis that a higher neoantigen load may enhance the immunogenicity of tumors, making them more susceptible to immune checkpoint inhibitors. This finding is corroborated by recent studies indicating that TMB can serve as a biomarker for immunotherapy response across various cancers, including melanoma and non-small cell lung cancer. However, our study also revealed that high TMB in the context of a high-risk score was associated with poorer prognosis, suggesting that the benefits of high TMB might be counteracted by other aggressive tumor features or a more suppressive immune environment. This complexity highlights the need for multifaceted models like ours that can dissect these interactions and provide more nuanced guidance for treatment planning. While our study represents a significant step forward in the prognostic modeling of ovarian cancer, it has several limitations. The retrospective nature of the data and reliance on public datasets, such as TCGA and GEO, may introduce selection biases and limit the generalizability of the findings to diverse patient populations, particularly due to incomplete clinical annotations, such as detailed treatment histories and co-morbidities. Furthermore, immune-related analyses based on in silico estimations may not fully capture the spatial and temporal heterogeneity of the tumor microenvironment, and the cross-sectional data used in this study do not account for dynamic changes in immune responses or tumor evolution over time. To address these limitations, future studies should incorporate longitudinal data and prospective cohorts to validate the model in diverse clinical settings, leveraging high-resolution technologies like single-cell sequencing and multiplex imaging to refine immune-related analyses. Additionally, expanding the model to include multi-omics data, such as proteomics and metabolomics, could enhance its predictive power and biological relevance. To ensure clinical applicability, a validation pathway should include external validation in independent cohorts, followed by prospective trials to evaluate its utility in stratifying patients for personalized therapies, such as immune checkpoint inhibitors or chemotherapy. Ultimately, integrating the model into routine practice would require the development of accessible tools, such as web-based platforms or clinical decision-support systems, to facilitate its adoption and enhance its impact on patient care. In conclusion, our study contributes to the evolving field of personalized medicine in ovarian cancer by providing a comprehensive prognostic model that integrates genetic, clinical, and immunological data. By offering insights into the complex interplay between the tumor and its microenvironment, our model holds the potential to guide more personalized and effective treatment strategies, ultimately improving the prognosis and quality of life for ovarian cancer patients.

Ovarian cancer is one of the most lethal gynecological malignancies globally, characterized by late-stage diagnosis and poor prognosis [ 1 – 3 ]. The complexity of ovarian cancer, subtle symptoms, and lack of effective early screening contribute to its high mortality rate [ 4 , 5 ]. This underscores the need for better diagnostic, prognostic, and therapeutic strategies to enhance patient outcomes and enable personalized treatments [ 6 ]. The pathogenesis of ovarian cancer is influenced by a multitude of factors, including genetic predispositions, environmental influences, and intricate molecular pathways [ 7 – 9 ]. Advancements in molecular biology and genomics have revealed significant heterogeneity in ovarian cancer, highlighting distinct molecular subtypes with different biological behaviors, treatment responses, and patient outcomes [ 10 , 11 ]. While molecular stratification has led to targeted therapies, ovarian cancer survival rates have only modestly improved over the past few decades [ 12 , 13 ]. This highlights the need for refined biomarkers to accurately predict disease progression, prognosis, and treatment response. In recent years, the role of the tumor microenvironment (TME) in the progression and metastasis of cancers, including ovarian cancer, has garnered significant attention [ 14 , 15 ]. The TME, including immune cells, fibroblasts, endothelial cells, and extracellular matrix, is crucial in cancer cell growth, therapy resistance, and immune evasion [ 16 – 18 ]. Immune cells within the TME are pivotal, correlating with patient outcomes and treatment responses, including immunotherapy [ 19 , 20 ]. Immune checkpoint inhibitors, which have revolutionized the treatment of several types of cancer, have shown promise in ovarian cancer, emphasizing the importance of understanding the immunological landscape of this malignancy [ 21 , 22 ]. Furthermore, the concept of the tumor mutational burden (TMB) has emerged as a potent predictor of response to immunotherapy [ 23 , 24 ]. High TMB is associated with increased neoantigen load, enhancing tumor immunogenicity and the efficacy of immune checkpoint blockade therapies [ 25 – 27 ]. However, the variability in TMB and its relationship with prognosis and treatment response in ovarian cancer remains underexplored. Given these complexities, there is a critical need for comprehensive prognostic models that can integrate molecular and immunological insights to predict patient outcomes more accurately. Such models could facilitate the identification of high-risk patients who might benefit from aggressive and tailored therapeutic strategies, potentially improving survival rates. This study aims to develop a prognostic model based on a risk score derived from gene expression profiles correlating with survival and treatment responses in ovarian cancer patients. The model aims to combine genetic, clinical, and immunological data to provide a robust tool for predicting patient outcomes. This is particularly crucial in a disease like ovarian cancer, where the selection of the most effective treatment strategy could significantly impact the patient’s quality of life and survival. To achieve these objectives, the study will utilize comprehensive genomic and clinical data from large cohorts, such as The Cancer Genome Atlas (TCGA), to identify key genes and pathways that are associated with ovarian cancer prognosis. These findings will be integrated into a risk scoring system that considers both the molecular characteristics of the tumors and the immune landscape, as indicated by immune cell infiltration and function within the TME. This dual approach ensures that the model captures the complex interactions between the tumor cells and their microenvironment, which are critical for the disease's progression and response to treatment. By focusing on these aspects, the research aims to contribute significantly to the personalized medicine landscape for ovarian cancer, providing clinicians with a valuable tool to tailor interventions based on individual patient risk profiles and enhancing our understanding of the disease’s underlying biology. This could ultimately lead to more effective management strategies, improved patient outcomes, and a reduction in the overall burden of ovarian cancer.