DOI:https://doi.org/10.65281/734509
Sirui Lü¹,Qiuying Yu¹, Lin Wang², Mingming Zheng³, Lei Chen³, Min Ye¹, Yang Hu¹, Wenxi Chen¹, Junxing Liu¹*
¹Jiamusi University, Jiamusi, Heilongjiang Province, P.R. 154002,China
²The First Affiliated Hospital of Jiamusi University, Jiamusi, Heilongjiang Province, P.R. 154002,China
³Jiamusi Infectious Disease Hospital, Jiamusi, Heilongjiang Province, P.R. 154002,China
*Corresponding author. Email: liujunxing@jmsu.edu.cn
Abstract
Background: Ovarian cancer (OV) is one of the most lethal gynecological malignancies worldwide. Migrasomes, newly identified migration-derived membranous organelles, play important roles in intercellular communication and regulation of the tumor microenvironment; however, their value in prognostic assessment of ovarian cancer remains unclear. This study aimed to construct a prognostic model based on migrasome-related long non-coding RNAs (lncRNAs) and to provide a new basis for precise risk stratification and individualized treatment in OV patients.
Methods: Clinical and transcriptomic data of OV patients were obtained from The Cancer Genome Atlas (TCGA) database. Prognostically significant migrasome-related lncRNAs were identified through co-expression analysis, univariate Cox regression, and least absolute shrinkage and selection operator (LASSO) regression. A prognostic lncRNA signature was established on the basis of known migrasome-related genes (MRGs). Kaplan–Meier survival analysis, time-dependent receiver operating characteristic (ROC) curves, concordance index (C-index), and principal component analysis (PCA) were used to evaluate model performance. Gene set enrichment analysis (GSEA), immune infiltration analysis (ssGSEA), tumor mutation burden (TMB) analysis, and drug sensitivity analysis based on the Genomics of Drug Sensitivity in Cancer (GDSC) database were further performed.
Results: A total of 15 migrasome-related lncRNAs associated with OV prognosis were identified, and a prognostic model stratifying patients into high-risk and low-risk groups was successfully established. Kaplan–Meier survival analysis showed that patients in the high-risk group had significantly poorer survival outcomes than those in the low-risk group. Univariate and multivariate analyses confirmed that the model could serve as an independent predictor of clinical outcomes. PCA demonstrated the stability of the model, while ROC curves and C-index analysis further validated its predictive performance. In addition, high-risk patients exhibited increased immune cell infiltration and lower tumor mutation burden, both of which were associated with unfavorable prognosis.
Conclusion: This study established, for the first time, a migrasome-related lncRNA prognostic model for ovarian cancer and revealed a potential mechanism by which the migrasome–lncRNA axis may influence tumor progression through regulation of the immune microenvironment. These findings provide new insights into disease progression and potential individualized therapeutic strategies.
Keywords: migrasome; ovarian cancer; lncRNA; immune infiltration; tumor mutation burden; drug sensitivity
- Introduction
Ovarian cancer (OV) is the deadliest malignancy of the female reproductive system and remains a major cause of cancer-related mortality worldwide [1,2,21]. Despite advances in surgical treatment, platinum-based chemotherapy, targeted therapy, and maintenance strategies, the prognosis of patients with ovarian cancer remains unsatisfactory, largely because more than 70% of patients are diagnosed at advanced stages and frequently develop recurrence or drug resistance [1,2,4]. High-grade serous ovarian carcinoma (HGSOC), the most common histological subtype, is characterized by nearly universal TP53 mutation and frequent homologous recombination deficiency, highlighting the molecular complexity of this disease [3]. Therefore, identification of reliable prognostic biomarkers and novel therapeutic targets is of considerable importance.
Clinically, ovarian cancer is not a single disease entity but a highly heterogeneous group of malignancies with marked differences in histology, genomic architecture, treatment response, and survival. High-grade serous ovarian carcinoma (HGSOC) accounts for the majority of deaths and is characterized by extensive chromosomal instability, widespread copy-number alteration, and a strong propensity for transcoelomic dissemination within the peritoneal cavity [1-4]. Even when cytoreductive surgery and platinum-based chemotherapy initially achieve remission, many patients later relapse with chemoresistant disease, reflecting substantial intratumoral heterogeneity and adaptive remodeling of the tumor microenvironment [2,4,24]. The clinical introduction of PARP inhibitors and anti-angiogenic therapy has improved outcomes in selected subgroups, yet durable benefit remains uneven and many patients still lack robust molecular markers that can assist prognostic assessment beyond conventional clinicopathological variables [2,4]. This unmet need provides a strong rationale for transcriptome-based biomarkers capable of capturing biological features related to dissemination, immune escape, and therapeutic vulnerability.
Migrasomes are a newly identified class of organelles generated on retraction fibers during cell migration and were first described by Ma et al. [6]. These membrane-bound structures are enriched in tetraspanins such as TSPAN4 and participate in the release of cytoplasmic contents, thereby mediating intercellular communication and signaling [5,6,13]. Increasing evidence suggests that migrasomes are involved in diverse physiological and pathological processes, including embryonic development, immune regulation, angiogenesis, and tumor progression [7,8,14]. In the tumor microenvironment, migrasomes may influence extracellular matrix remodeling, cell migration, and immune cell function, suggesting that migrasome-related pathways may be closely associated with cancer progression [7,8].
At the biological level, migrasome formation depends on cell migration, retraction fibers, membrane tension, and the coordinated action of tetraspanins and adhesion-related molecules such as TSPAN4, integrins, and associated cytoskeletal regulators [5,6,13,22]. This makes migrasomes conceptually relevant to malignant tumors with strong migratory and invasive capacity. In ovarian cancer, tumor cells continuously interact with mesothelial surfaces, stromal matrices, ascitic fluid, and immune cells during intraperitoneal spread; each of these contexts could plausibly favor migrasome-mediated communication. Although direct experimental evidence in ovarian cancer is still limited, current knowledge from other systems supports the hypothesis that migrasomes may contribute to metastatic fitness by coordinating matrix interaction, signal transfer, and local microenvironmental conditioning [22,23].
Long non-coding RNAs (lncRNAs) are transcripts longer than 200 nucleotides that play important roles in gene regulation, chromatin remodeling, and cancer biology [9–11]. Although most lncRNAs remain insufficiently characterized, many have been shown to participate in tumor proliferation, invasion, metastasis, immune escape, and therapeutic resistance [9–11]. However, the roles of migrasome-related lncRNAs in ovarian cancer have not been systematically investigated.
Beyond their original description as migration-dependent membranous structures, migrasomes are increasingly regarded as active regulators of cancer-associated communication rather than passive by-products of cell movement. Recent reviews have proposed that migrasomes may participate in extracellular matrix remodeling, angiogenesis, metabolic adaptation, and immune regulation, while pan-cancer analyses further suggest that the migrasome regulator TSPAN4 is associated with suppressive immune phenotypes and macrophage polarization [22,23]. These observations are particularly relevant to ovarian cancer, in which stromal remodeling and an immunosuppressive microenvironment are major determinants of metastasis, therapeutic resistance, and clinical outcome [24].
In parallel, lncRNAs have emerged as key regulators of ovarian cancer cell proliferation, metastatic dissemination, chemotherapy resistance, and tumor–immune crosstalk, and recent reviews have emphasized their value as diagnostic, prognostic, and therapeutic biomarkers [25]. Notably, ovarian-cancer-specific studies have shown that lncRNAs can directly influence immune escape through mechanisms involving PD-L1 regulation and crosstalk with macrophages [35,36]. Therefore, a prognostic framework based on migrasome-related lncRNAs is conceptually attractive because it may capture both migration-associated intercellular communication and transcriptional programs linked to immune adaptation. However, the intersection between migrasome biology and lncRNA regulation in ovarian cancer has not yet been systematically characterized, which provides the rationale for the present study [22,25].
From the perspective of biomarker development, the intersection between migrasome biology and lncRNA regulation is especially attractive. Migrasomes arise in the setting of active cell migration and can carry proteins, RNAs, and signaling molecules that influence neighboring cells or the extracellular environment [6,13,22]. lncRNAs, in turn, are highly context-dependent transcripts that often reflect cell-state transitions, stress adaptation, and tumor-microenvironment crosstalk more sensitively than protein-coding genes alone [9-11,25]. A prognostic framework built on migrasome-related lncRNAs may therefore serve as a surrogate readout of migratory activity, intercellular communication, and immune remodeling. In ovarian cancer, where dissemination through ascites, matrix interaction, and stromal conditioning are central to disease progression, such an approach may provide a biologically meaningful complement to traditional survival models [22-25].
In the present study, we constructed a prognostic signature based on migrasome-related lncRNAs using TCGA-OV data and comprehensively assessed its prognostic performance, biological significance, immune landscape, mutation characteristics, and drug sensitivity associations. Our findings may provide new insights into the interplay between migrasome biology and ovarian cancer progression and may support the development of more precise prognostic and therapeutic strategies.
- Materials and Methods
2.1 Data acquisition and screening of migrasome-related lncRNAs
RNA sequencing data, somatic mutation data, and corresponding clinical information of ovarian cancer patients were downloaded from The Cancer Genome Atlas (TCGA) database [3,12]. Perl software (version 5.30.0.1) was used to extract mRNA expression data, lncRNA expression data, mutation matrices, and clinical information. Samples with incomplete clinicopathological data were excluded from subsequent analyses.
Migrasome-related genes (MRGs) were identified through literature review and GeneCards database screening using the keyword “migrasome” with a relevance score greater than 1 [13–15]. Ultimately, 22 MRGs were included in this study: ITGB1, ITGA5, EOGT, CPQ, PIGK, NDST1, TSPAN4, EPCIP, PKD2, PKD1, ESR1, FN1, KDM1A, ROCK1, ATF6, RAB10, SYT1, PAK4, CXCL12, LAMA4, PLA2R1, and CD151. Pearson correlation analysis was then performed to identify migrasome-related lncRNAs based on co-expression relationships between MRGs and lncRNAs, with the thresholds set at |r| > 0.4 and P < 0.001. The “limma” R package was used for normalization and downstream analysis [16].
2.2 Construction and validation of the prognostic model
The entire OV cohort was randomly divided into a training set and a testing set at a 1:1 ratio. In the training set, univariate Cox regression analysis was first performed to identify prognostic migrasome-related lncRNAs. Candidate lncRNAs with P < 0.05 were then subjected to least absolute shrinkage and selection operator (LASSO) regression to minimize overfitting and retain the most informative variables [17,18,32,33]. Finally, multivariate Cox regression analysis was used to establish the prognostic model and calculate the regression coefficients of the selected lncRNAs.
To improve model robustness and reduce the risk of overfitting, we adopted a stepwise feature-selection strategy combining univariate screening, penalized regression, and multivariate Cox modeling. This workflow has practical advantages in high-dimensional transcriptomic studies because it progressively narrows the candidate space while retaining interpretability and clinical usability [17,18,32,33]. The use of a training/testing split allows internal validation, whereas Kaplan-Meier analysis, ROC curves, C-index evaluation, PCA, and nomogram construction together provide complementary information on discrimination, calibration, and clinical applicability. The median risk score was selected as the cutoff to facilitate reproducible patient stratification and avoid an overly optimized threshold that might not generalize well in future datasets.
The risk score for each patient was calculated as a linear combination of lncRNA expression levels weighted by their corresponding coefficients. Patients were classified into high-risk and low-risk groups according to the median risk score of the training cohort. Kaplan–Meier survival analysis was used to compare overall survival (OS) and progression-free survival (PFS) between the two groups. Univariate and multivariate Cox analyses were performed to evaluate whether the risk score could serve as an independent prognostic factor. Time-dependent ROC analysis, concordance index (C-index), principal component analysis (PCA), nomogram construction, and calibration analysis were used to evaluate the predictive ability and clinical utility of the model.
2.3 Functional enrichment, immune infiltration, mutation, and drug sensitivity analyses
Differentially expressed genes (DEGs) between the high-risk and low-risk groups were identified with the criteria |logFC| > 1 and false discovery rate (FDR) < 0.05. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were conducted using the “clusterProfiler” package [19]. Gene set enrichment analysis (GSEA) was additionally performed to explore the biological pathways differentially enriched between the two groups [31].
To characterize the tumor immune microenvironment, the “ESTIMATE” algorithm was used to calculate stromal, immune, and ESTIMATE scores [26]. Immune cell infiltration was assessed using CIBERSORT [27], and single-sample gene set enrichment analysis (ssGSEA) implemented in the “GSVA” package was used to evaluate immune cell composition and immune-related functions [20]. Somatic mutation data in MAF format were used for tumor mutation burden (TMB) analysis, and mutation landscapes were visualized with the “maftools” package [28]. TIDE analysis was performed to estimate the potential response to immunotherapy [29]. Drug sensitivity analysis was conducted using the Genomics of Drug Sensitivity in Cancer (GDSC) database [34], and the “oncoPredict” package was used to estimate half-maximal inhibitory concentration (IC50) values for candidate compounds [30].
- Results
3.1 Identification of prognostic migrasome-related lncRNAs and establishment of the signature
A total of 1,301 migrasome-related lncRNAs were identified through co-expression analysis between lncRNAs and the 22 MRGs. Univariate Cox regression analysis further identified 30 migrasome-related lncRNAs significantly associated with ovarian cancer prognosis. After LASSO regression and multivariate Cox regression, 15 lncRNAs were finally selected to construct the prognostic signature: MIR600HG, AC010615.2, AC007598.1, SUCLG2-AS1, TRAM2-AS1, AC046158.2, AC027130.1, LINC02798, AC106801.1, AC019080.1, AC053503.3, AL355999.1, PTPRD-AS1, AC011455.8, and AC109587.1. As shown in Figure 1, co-expression analysis identified migrasome-related lncRNAs associated with 22 migrasome-related genes, univariate Cox regression screened prognosis-related candidates, and LASSO regression further reduced the variables to generate the optimal lncRNA set for model construction.
Figure 1. Identification of prognostic migrasome-related lncRNAs in ovarian cancer.
(A) Sankey diagram showing the co-expression relationships between migrasome-related genes and lncRNAs. (B) Forest plot of univariate Cox regression analysis for prognostic migrasome-related lncRNAs. (C) Cross-validation curve for determining the optimal penalty parameter (λ) in the LASSO regression model. (D) LASSO coefficient profiles of candidate migrasome-related lncRNAs.
The risk score was calculated as follows:
Risk score =
(0.3243 × MIR600HG) + (−0.2269 × AC010615.2) + (1.2634 × AC007598.1) + (−0.9809 × SUCLG2-AS1) + (−0.6182 × TRAM2-AS1) + (−1.0211 × AC046158.2) + (−0.8661 × AC027130.1) + (0.8151 × LINC02798) + (−2.0596 × AC106801.1) + (1.1845 × AC019080.1) + (0.8893 × AC053503.3) + (−0.2329 × AL355999.1) + (0.3381 × PTPRD-AS1) + (0.5294 × AC011455.8) + (−1.0553 × AC109587.1).
Patients were subsequently divided into high-risk and low-risk groups based on the median risk score. Baseline clinical characteristics did not differ significantly between the training and testing cohorts, suggesting that the random grouping strategy was balanced and reliable.As shown in Figure 2, the correlations between migrasome-related genes and the 15 model lncRNAs were visualized, and Kaplan–Meier analysis demonstrated significantly poorer overall survival and progression-free survival in the high-risk group than in the low-risk group. Risk distribution, survival status, and expression heatmap analyses further illustrated the prognostic stratification ability of the signature.
Figure 2. Construction and visualization of the migrasome-related lncRNA prognostic signature in ovarian cancer.
(A) Heatmap showing the correlations between migrasome-related genes and the 15 model-associated lncRNAs. (B, C) Kaplan–Meier curves for overall survival (OS) and progression-free survival (PFS) in the high-risk and low-risk groups. (D) Distribution of risk scores among all OV patients. (E) Survival status scatter plot of patients ranked by increasing risk score. (F) Heatmap of the expression profiles of the 15 prognostic migrasome-related lncRNAs in the two risk groups.
The baseline clinical characteristics of patients in the training and testing cohorts are summarized in Table 1. No significant differences were observed between the two cohorts with respect to age, gender, or tumor grade, indicating that the random grouping was appropriate and comparable.
Table 1. Baseline clinical characteristics of ovarian cancer patients in the training and testing cohorts.
| Covariates | Type | Total N=425 | Test N1=212 | Train N2=213 | P-value |
| Age | <=65 | 292(68.71%) | 151(71.23%) | 141(66.2%) | 0.3109 |
| Age | >65 | 133(31.29%) | 61(28.77%) | 72(33.8%) | |
| Gender | FEMALE | 425(100%) | 212(100%) | 213(100%) | 0.9613 |
| Grade | G1 | 1(0.24%) | 1(0.47%) | 0(0%) | 0.5685 |
| Grade | G2 | 47(11.06%) | 23(10.85%) | 24(11.27%) | |
| Grade | G3 | 368(86.59%) | 184(86.79%) | 184(86.38%) | |
| Grade | G4 | 1(0.24%) | 0(0%) | 1(0.47%) | |
| Grade | unknow | 8(1.88%) | 4(1.89%) | 4(1.88%) |
In the training cohort, the risk distribution plot, survival status plot, and heatmap showed a negative association between survival time and risk score. A similar trend was also observed in the testing cohort. Kaplan–Meier analysis demonstrated that the high-risk group had significantly poorer overall survival than the low-risk group in both cohorts (Figure 3).
The composition of the final model suggests that ovarian cancer prognosis is shaped by a balance of both risk-promoting and risk-protective transcriptional programs. Positive regression coefficients may reflect lncRNAs associated with aggressive biological behavior, stromal activation, or immune evasion, whereas negative coefficients may indicate transcripts linked to relatively restrained tumor behavior or preserved regulatory homeostasis. Such mixed-direction signatures are common in transcriptomic prognostic models and biologically plausible because survival is determined not by a single pathway but by the integrated effects of tumor-intrinsic aggressiveness, microenvironmental adaptation, and host response. Therefore, the 15-lncRNA panel may be interpreted as a systems-level summary of migrasome-related tumor states rather than a set of isolated markers.
Figure 3. Validation of the migrasome-related lncRNA prognostic signature in the training and testing cohorts.
(A–C) Risk score distribution, survival status, and heatmap of the 15 prognostic migrasome-related lncRNAs in the training cohort. (D–F) Risk score distribution, survival status, and heatmap of the 15 prognostic migrasome-related lncRNAs in the testing cohort. (G, H) Kaplan–Meier overall survival curves for high-risk and low-risk patients in the training and testing cohorts, respectively.
3.2 Prognostic performance and independent predictive value of the model
Kaplan–Meier survival analysis demonstrated that patients in the high-risk group had significantly worse overall survival and progression-free survival than those in the low-risk group. Risk distribution plots, survival status plots, and heatmaps of lncRNA expression further indicated that higher risk scores were associated with more deaths and shorter survival times.
In both the training and testing sets, the prognostic signature consistently distinguished patients with poor survival outcomes from those with more favorable prognosis. Univariate Cox regression analysis showed that age and risk score were significantly associated with prognosis, while multivariate Cox regression confirmed that the risk score remained an independent prognostic factor for ovarian cancer. These findings indicate that the migrasome-related lncRNA signature has robust and reproducible prognostic value.
Time-dependent ROC analysis showed that the area under the curve (AUC) values for 1-, 3-, and 5-year overall survival were 0.748, 0.726, and 0.737, respectively, indicating favorable predictive accuracy. C-index analysis further suggested that the prognostic performance of the risk score was superior to conventional clinical variables. A nomogram integrating age, grade, and risk score was constructed to improve clinical applicability, and calibration curves showed good agreement between predicted and observed survival probabilities.As shown in Figure 4, univariate and multivariate Cox regression analyses confirmed that the risk score was an independent prognostic factor for ovarian cancer. ROC and C-index analyses demonstrated that the signature had superior predictive performance compared with conventional clinical variables, while the nomogram and calibration curves further supported its clinical applicability.
These performance data indicate that the prognostic signature adds clinically relevant information beyond routine variables. In TCGA-OV, most tumors are high grade, which limits the discriminatory value of grade alone; by contrast, transcriptomic risk scores can capture latent biological diversity that is not readily visible through conventional clinical descriptors. The favorable AUC and C-index values, together with the nomogram results, suggest that the migrasome-related lncRNA model may be useful as an adjunct tool for postoperative risk stratification, survival estimation, and potentially for identifying patients who require closer surveillance or more individualized therapeutic planning.
Figure 4. Independent prognostic analysis and predictive performance evaluation of the migrasome-related lncRNA signature in ovarian cancer.
(A, B) Univariate and multivariate Cox regression analyses of clinicopathological variables and risk score. (C) ROC curves comparing the predictive performance of clinical variables and risk score. (D) C-index curves of different prognostic factors. (E) Time-dependent ROC curves for predicting 1-, 3-, and 5-year overall survival. (F) Nomogram integrating age, grade, and risk score for predicting survival probability in OV patients. (G) Calibration curves of the nomogram for 1-, 3-, and 5-year survival prediction.
3.3 Clinical subgroup analysis and principal component analysis
To determine whether the model remained stable across different clinical backgrounds, age-stratified survival analysis was performed. The results showed that the prognostic signature could effectively distinguish high-risk from low-risk patients in both the ≤65-year and >65-year subgroups, indicating that the model retained predictive value across age categories.
Principal component analysis was conducted based on all genes, all migrasome-related genes, all migrasome-related lncRNAs, and the 15-lncRNA signature. Compared with the other datasets, the 15-lncRNA model showed a much clearer separation between high-risk and low-risk patients, suggesting that this signature had strong discriminatory capacity and superior subgrouping performance.As shown in Figure 5, the prognostic signature remained effective in both age subgroups, with high-risk patients showing significantly poorer overall survival than low-risk patients. PCA further demonstrated that the 15-lncRNA signature provided a clearer separation between the two risk groups than analyses based on all genes, migrasome-related genes, or all migrasome-related lncRNAs.
Figure 5. Clinical subgroup analysis and principal component analysis of the migrasome-related lncRNA prognostic signature in ovarian cancer.
(A, B) Kaplan–Meier overall survival curves of high-risk and low-risk patients stratified by age (≤65 years and >65 years). (C) PCA plot based on the expression profiles of all genes. (D) PCA plot based on migrasome-related genes. (E) PCA plot based on migrasome-related lncRNAs. (F) PCA plot based on the 15 model-associated migrasome-related lncRNAs.
3.4 Functional enrichment analysis of the risk groups
To explore the biological mechanisms underlying the observed prognostic differences, GO, KEGG, and GSEA analyses were performed on the DEGs between the high-risk and low-risk groups. GO enrichment analysis revealed that these genes were mainly associated with nucleosome assembly, neuropeptide signaling pathways, collagen-containing extracellular matrix, perinuclear regions, chromatin structural constituents, and neuropeptide Y receptor activity. KEGG pathway analysis indicated significant enrichment in viral carcinogenesis-related pathways and neutrophil-associated signaling pathways.
GSEA showed that genes in the high-risk group were enriched mainly in extracellular matrix organization, structural remodeling, and ossification-related pathways, whereas those in the low-risk group were enriched in pathways related to nuclear assembly and genetic material regulation. These findings suggest that the two risk groups exhibit distinct biological characteristics and may reflect different states of tumor aggressiveness and microenvironmental interaction.As shown in Figure 6, GO and KEGG enrichment analyses revealed distinct functional characteristics between the high-risk and low-risk groups, while GSEA further demonstrated that high-risk tumors were mainly enriched in extracellular matrix and structural remodeling pathways, whereas low-risk tumors were enriched in pathways related to nuclear assembly and genetic material regulation.
The enrichment results are consistent with the invasive biology of advanced ovarian cancer. Processes such as extracellular matrix organization, collagen remodeling, and structural reprogramming are central to peritoneal implantation, mesothelial adhesion, and metastatic colonization. Accordingly, the pathways enriched in the high-risk group support the interpretation that the risk signature captures a more motile and microenvironmentally interactive tumor phenotype. In contrast, the comparatively different enrichment pattern in the low-risk group suggests that these tumors may preserve alternative regulatory states less dominated by stromal remodeling and tissue invasion [22,24,25].
Figure 6. Functional enrichment analysis of differentially expressed genes between the high-risk and low-risk groups in ovarian cancer.
(A) Circular plot of GO enrichment analysis, including biological process (BP), cellular component (CC), and molecular function (MF) categories. (B, C) Bubble plot and bar plot of significantly enriched GO terms. (D, E) Bar plot and bubble plot of significantly enriched KEGG pathways. (F) GSEA plot showing representative pathways enriched in the high-risk group. (G) GSEA plot showing representative pathways enriched in the low-risk group.
3.5 Immune infiltration characteristics associated with the prognostic signature
To investigate the relationship between the risk model and the tumor immune microenvironment, ESTIMATE, CIBERSORT, and ssGSEA analyses were performed. The high-risk group had significantly higher ESTIMATE scores and immune-related scores than the low-risk group, while tumor purity was relatively lower, suggesting a positive relationship between the risk score and immune infiltration.
CIBERSORT analysis showed significant differences in immune cell composition between the two groups. Specifically, resting memory CD4 T cells, follicular helper T cells, and M1 macrophages were significantly more abundant in the low-risk group. In contrast, ssGSEA results showed that macrophages, mast cells, and T helper cells were significantly enriched in the high-risk group. These results suggest that the migrasome-related lncRNA signature is closely associated with immune microenvironment remodeling in ovarian cancer.As shown in Figure 7, the high-risk group exhibited higher ESTIMATE-related scores and lower tumor purity, indicating a closer association between risk score and immune infiltration. CIBERSORT and ssGSEA analyses further revealed significant differences in immune cell composition and immune-related functions between the high-risk and low-risk groups.
Taken together, the immune analyses suggest that the high-risk group is characterized by quantitatively greater but functionally less effective immune infiltration. In ovarian cancer, abundant immune or stromal content does not necessarily translate into active antitumor immunity, because the infiltrate may be enriched for suppressive myeloid populations, dysfunctional T-cell programs, or checkpoint-associated inhibitory states. The observed changes in macrophages, helper T-cell-associated signatures, and ESTIMATE-derived scores therefore support the view that the migrasome-related lncRNA model reflects immune remodeling and immunosuppressive contexture rather than simple immune activation [23,24,35,36].
Figure 7. Association between the migrasome-related lncRNA prognostic signature and the immune microenvironment in ovarian cancer.
(A) Comparison of stromal score, immune score, ESTIMATE score, and tumor purity between the high-risk and low-risk groups. (B) Boxplots showing differences in the infiltration of 22 immune cell types between the two risk groups based on CIBERSORT analysis. (C) Stacked bar plot showing the relative abundance of immune cell infiltration in individual ovarian cancer samples from the low-risk and high-risk groups. (D) Boxplots of ssGSEA scores showing differences in immune-related functions between the high-risk and low-risk groups.
3.6 Tumor mutation burden and mutation landscape analysis
Tumor mutation burden analysis demonstrated that the low-risk group had significantly higher TMB scores than the high-risk group. TIDE analysis suggested that patients in the high-risk group might derive greater benefit from immunotherapy. Mutation landscape analysis revealed that TP53 and TTN were among the most frequently mutated genes in both risk groups, with mutation frequencies exceeding 25%.
Further survival analysis based on TMB status showed that patients with high TMB had better overall survival than those with low TMB. When risk group and TMB status were combined, the high-TMB/low-risk subgroup exhibited the best prognosis, whereas the low-TMB/high-risk subgroup showed the worst survival outcome. These findings indicate that the prognostic signature and mutation burden jointly contribute to prognostic stratification in OV.As shown in Figure 8, the low-risk group exhibited a significantly higher TMB than the high-risk group, whereas TIDE analysis suggested a potentially better immunotherapy response in the high-risk group. Mutation landscape and survival analyses further demonstrated that TMB status, especially when combined with risk stratification, was closely associated with clinical outcome in OV patients.
The joint analysis of TMB and risk score further suggests that these variables are complementary rather than redundant. TMB provides a partial estimate of mutational antigenicity, whereas the lncRNA-based risk signature appears to capture additional dimensions of tumor biology, including stromal context, migratory behavior, and immune dysfunction. This may explain why patients with similar mutational burdens can nonetheless display different survival outcomes and distinct predicted responses to immunotherapy. Combining genomic and transcriptomic indicators may therefore offer a more informative framework for precision stratification in ovarian cancer [24,29].
Figure 8. Tumor mutation burden, immunotherapy response prediction, and mutation landscape analysis in ovarian cancer.
(A) Comparison of tumor mutation burden (TMB) scores between the high-risk and low-risk groups. (B) TIDE analysis comparing the predicted immunotherapy response between the two risk groups. (C) Waterfall plot showing the mutation landscape of the high-risk group. (D) Waterfall plot showing the mutation landscape of the low-risk group. (E) Kaplan–Meier survival curve comparing overall survival between the high-TMB and low-TMB groups. (F) Kaplan–Meier survival curves of four subgroups stratified by both TMB status and risk group.
3.7 Drug sensitivity analysis
Drug sensitivity analysis was performed to further investigate the potential therapeutic implications of the migrasome-related lncRNA signature. The results showed that the high-risk group appeared to be more sensitive to BMS-536924, BMS-754807, Foretinib, and Pictilisib, whereas the low-risk group was more sensitive to Axitinib, Cyclophosphamide, Dihydrorotenone, and Gallibiscoquinazole. These findings suggest that the risk model may have value not only in prognosis prediction but also in guiding individualized therapeutic selection for ovarian cancer patients.As shown in Figure 9, the high-risk and low-risk groups displayed distinct predicted drug sensitivity profiles. High-risk patients appeared to be more sensitive to BMS-536924, BMS-754807, Foretinib, and Pictilisib, whereas low-risk patients were more sensitive to Axitinib, Cyclophosphamide, Dihydrorotenone, and Gallibiscoquinazole.
Although exploratory in nature, the drug sensitivity results raise the possibility that migrasome-related transcriptional states are accompanied by pathway-specific therapeutic vulnerabilities. Such predictions cannot be interpreted as direct evidence of clinical efficacy, but they are useful for prioritizing agents for further functional testing in cell lines, organoids, or patient-derived models. If validated experimentally, the differential IC50 patterns observed here may help identify candidate compounds or rational combinations for risk-defined ovarian cancer subgroups [30,34].
Figure 9. Predicted drug sensitivity analysis of the migrasome-related lncRNA prognostic signature in ovarian cancer.
(A–D) Boxplots showing that the high-risk group is more sensitive to BMS-536924, BMS-754807, Foretinib, and Pictilisib. (E–H) Boxplots showing that the low-risk group is more sensitive to Axitinib, Cyclophosphamide, Dihydrorotenone, and Gallibiscoquinazole. Drug sensitivity was evaluated based on the estimated IC50 values.
- Discussion
Ovarian cancer remains one of the most challenging gynecologic malignancies because of its high mortality, delayed diagnosis, and frequent development of recurrence and chemoresistance [1,2,4,21]. Although molecular classification and targeted therapy have improved the management of some patients, there is still a pressing need for more effective prognostic biomarkers and treatment stratification tools. In the present study, we established a novel migrasome-related lncRNA signature that effectively stratified ovarian cancer patients into distinct prognostic groups and provided insight into their immune and mutational characteristics.
Migrasomes are increasingly recognized as key mediators of cell migration-associated communication and have been implicated in tumor biology, angiogenesis, and immune regulation [5–8,13,14]. However, their clinical significance in ovarian cancer has remained largely unexplored. By integrating migrasome biology with lncRNA-based modeling, our study extends the current understanding of how migration-related organelles may contribute to tumor progression and patient outcome.
From a mechanistic perspective, the present model may reflect the ability of migrasome-associated programs to reshape intercellular communication within the ovarian cancer microenvironment. Migrasomes can package signaling molecules and are increasingly linked to matrix remodeling, angiogenesis, and immune modulation, whereas ovarian cancer progression is strongly influenced by immunosuppressive myeloid populations, dysfunctional T-cell states, and stromal interactions [22–24]. Thus, the association between the risk score and immune infiltration observed in our study is biologically plausible rather than merely statistical.
An important contribution of the present work is that it connects two research fields that have so far largely developed in parallel: migrasome biology and lncRNA-based prognostic modeling. Most previously reported ovarian cancer signatures have been built around immune genes, metabolism-related pathways, autophagy, ferroptosis, or other cell-death programs, whereas the present study uses migration-associated organelles as the biological anchor of model construction. This conceptual difference matters because ovarian cancer progression is driven not only by proliferation but also by dissemination through the peritoneal cavity, dynamic interaction with mesothelial and stromal compartments, and adaptation to immune pressure. A migrasome-centered lncRNA model may therefore be particularly well suited to capture clinically relevant aspects of ovarian cancer aggressiveness [22,24,25].
The 15-lncRNA signature identified in this study showed stable prognostic value in both the training and testing cohorts and remained an independent prognostic factor after adjustment for conventional clinical variables. These results suggest that migrasome-related lncRNAs may reflect important biological features that are not fully captured by routine clinicopathological indicators. In addition, PCA, ROC, C-index, and nomogram analyses collectively demonstrated the robustness and clinical applicability of the model.
Another notable feature of the signature is that many of the included lncRNAs remain poorly characterized in ovarian cancer. This can be seen as both a limitation and a scientific opportunity. On the one hand, the current study cannot determine whether each transcript directly regulates migrasome biogenesis, cargo sorting, immune signaling, or therapy resistance. On the other hand, unbiased transcriptomic modeling may reveal biologically important lncRNAs that would be missed by candidate-gene approaches focused only on well-studied molecules. Accordingly, the present signature should be viewed not only as a prognostic tool but also as a resource for future functional screening aimed at clarifying how specific lncRNAs influence migratory communication and tumor-microenvironment crosstalk [9-11,25].
Functional enrichment analyses suggested that high-risk tumors were associated mainly with extracellular matrix organization, structural remodeling, and ossification-related processes. These pathways are often linked to enhanced invasiveness, stromal interaction, and metastatic capacity. By contrast, low-risk tumors were enriched in pathways related to nuclear assembly and genetic regulation, implying relatively different biological programs. These findings indicate that the migrasome-related lncRNA signature may reflect not only prognosis but also fundamental biological heterogeneity in ovarian cancer.
The enrichment of extracellular matrix organization and structural remodeling in the high-risk group is also noteworthy. Ovarian cancer dissemination typically occurs through transcoelomic spread, a process highly dependent on cell motility, mesothelial adhesion, and matrix remodeling; therefore, migrasome-related signatures may be particularly well suited to capture aggressive migratory phenotypes [22,24]. Because lncRNAs are now known to regulate epithelial–mesenchymal plasticity, invasion, therapy resistance, and immune escape in ovarian cancer, the current signature may integrate migration-related vesicular biology with transcriptional programs that promote metastatic competence [25,35,36].
This interpretation is also consistent with the clinical pattern of ovarian cancer spread. Unlike many solid tumors that metastasize predominantly through the bloodstream, ovarian cancer commonly disseminates across the peritoneal cavity, where successful implantation depends on cell detachment, survival in suspension, adhesion to mesothelial surfaces, extracellular matrix remodeling, and subsequent colonization of secondary sites. A molecular program linked to migrasome biology may therefore be especially informative in this disease setting, because it is directly connected to migration-associated processes and intercellular communication during metastatic seeding. The functional enrichment pattern observed here fits well with that clinical reality and strengthens the biological credibility of the signature [2,24].
The immune analyses further reinforced the biological relevance of the signature. High-risk tumors displayed stronger immune-related scores and distinct immune cell infiltration patterns, suggesting that the migrasome-lncRNA axis may participate in remodeling of the tumor microenvironment. Interestingly, the low-risk group showed higher TMB, whereas TIDE analysis suggested that the high-risk group might potentially derive more benefit from immunotherapy. This apparent complexity reflects the multifaceted interaction between mutation burden, immune evasion, and tumor microenvironment status, and indicates that combined evaluation of risk score, immune infiltration, and mutation burden may provide more informative stratification than any single parameter alone.
The apparent coexistence of higher immune/stromal scores with poorer survival in the high-risk group should be interpreted cautiously. In ovarian cancer, increased immune infiltration does not necessarily indicate effective antitumor immunity, because the immune contexture may be dominated by suppressive macrophages, dysfunctional helper T-cell programs, exhausted T cells, or checkpoint-driven inhibitory states [24]. In this setting, the observed differences in TMB and TIDE may reflect distinct forms of immune escape rather than a simple “hot-versus-cold” tumor dichotomy [24,29].
The immune findings also have potential translational implications. Rather than indicating that all high-risk tumors are uniformly responsive to immunotherapy, the data suggest that different risk-defined subgroups may evade immune control through distinct mechanisms. Tumors enriched for stromal and macrophage-associated programs may require combination strategies that address checkpoint signaling together with myeloid reprogramming, angiogenesis, or matrix remodeling, whereas tumors with relatively higher TMB but less overt immunosuppressive signaling may follow a different therapeutic logic. In this sense, the risk model may eventually help refine patient selection for combination regimens instead of serving as a stand-alone predictor of response [24,29,35,36].
Drug sensitivity analysis revealed different predicted responses to multiple compounds between the two risk groups, suggesting that this signature may also have practical value in treatment selection. Although these findings are based on in silico prediction and require experimental validation, they provide an initial basis for exploring personalized treatment strategies in ovarian cancer.
Similarly, the drug sensitivity results should be regarded as hypothesis-generating rather than practice-changing. Computational platforms such as oncoPredict provide a practical framework for prioritizing candidate compounds using transcriptomic similarity to pharmacogenomic reference datasets, but these predictions cannot substitute for functional validation in ovarian cancer cell lines, organoids, or patient-derived models [30,34]. Even so, the differential sensitivity patterns observed here may provide a starting point for future studies exploring whether migrasome-related lncRNA states are linked to pathway-specific vulnerabilities.
Several limitations of this study should be acknowledged. First, the model was constructed and validated using retrospective TCGA data, and external validation in independent cohorts is still needed. Second, the biological mechanisms of the identified lncRNAs were not experimentally investigated. Third, the immunotherapy and drug sensitivity findings were based on computational prediction rather than clinical response data. Nevertheless, our study provides a novel perspective on the prognostic role of migrasome-related lncRNAs in ovarian cancer and offers a potentially valuable framework for future mechanistic and translational research.
Several additional methodological issues also merit attention. Because the model was derived from a single public cohort, hidden batch effects, incomplete clinical annotation, and cohort-specific sampling bias may have influenced feature selection and effect estimates. Moreover, although LASSO-based signatures are widely used in transcriptomic prognostic modeling, their stability can vary with sample composition and parameter tuning, making external validation and experimental confirmation essential before any clinical translation is attempted [32,33].
Future work should also focus on translating the present findings into experimentally testable hypotheses. For example, gain- and loss-of-function studies could determine whether selected lncRNAs alter the abundance, size, or cargo composition of migrasomes, while co-culture assays with macrophages or T cells could clarify whether these transcripts modulate immune-cell recruitment or polarization through migrasome-dependent mechanisms. Parallel evaluation in ascites samples, extracellular vesicle fractions, and spatial transcriptomic datasets may further help localize where the migrasome-lncRNA axis operates in the ovarian cancer ecosystem. Such studies would substantially strengthen the mechanistic foundation of the current prognostic observations [22-25,35,36].
Before any clinical application can be considered, several translational steps are required. External validation in independent public cohorts and in locally collected clinical samples is necessary to assess platform robustness and geographic generalizability. In addition, selected lncRNAs with strong coefficients or central network positions should be experimentally tested in ovarian cancer cell lines, co-culture systems, and ascites-derived models to determine whether they truly regulate migrasome formation, immune communication, or drug response. Finally, prospective studies should evaluate whether integrating the risk score with established clinical and molecular variables, such as BRCA or homologous recombination deficiency status, residual disease, and treatment response, improves real-world decision-making [3,4,24,25].
- Conclusion
In summary, we developed and validated a novel prognostic signature composed of 15 migrasome-related lncRNAs for ovarian cancer. This signature effectively stratified patients into high-risk and low-risk groups, demonstrated stable predictive performance across different analytical settings, and remained an independent predictor of survival outcomes. More importantly, beyond its prognostic value, the model was closely associated with several key biological and clinical characteristics, including immune cell infiltration, tumor mutation burden, and differential drug sensitivity. These findings suggest that migrasome-related lncRNAs may play an important role in ovarian cancer progression, possibly through modulation of the tumor immune microenvironment and related molecular pathways. The present study therefore provides new evidence supporting the potential involvement of the migrasome–lncRNA axis in ovarian cancer biology and highlights its possible utility in prognostic evaluation and personalized therapeutic decision-making. At the same time, this work offers a new perspective for understanding the molecular heterogeneity of ovarian cancer and may contribute to the identification of clinically relevant biomarkers and therapeutic targets. Nevertheless, because the present study was based primarily on retrospective public database analysis, further experimental validation, external cohort verification, and multicenter clinical studies are still needed to confirm the robustness, generalizability, and clinical applicability of this prognostic model, as well as to clarify the underlying biological mechanisms of the identified lncRNAs.
From a broader translational standpoint, the present work highlights the value of integrating emerging cell-biology concepts with transcriptomic risk modeling in ovarian cancer. Rather than treating prognostic signatures as purely statistical classifiers, anchoring model construction in a biologically coherent process such as migrasome-associated communication may improve interpretability and facilitate downstream mechanistic validation. If future studies confirm these findings in external cohorts and experimental systems, migrasome-related lncRNAs may become useful not only for prognosis prediction but also for patient stratification in studies of immunotherapy, microenvironment-targeted treatment, and metastasis-directed intervention.
References
[1] Siegel RL, Miller KD, Wagle NS, Jemal A. Cancer statistics, 2023. CA Cancer J Clin. 2023;73(1):17-48. doi:10.3322/caac.21763
[2] Lheureux S, Gourley C, Vergote I, Oza AM. Epithelial ovarian cancer. Lancet. 2019;393(10177):1240-1253. doi:10.1016/S0140-6736(18)32552-2
[3] The Cancer Genome Atlas Research Network. Integrated genomic analyses of ovarian carcinoma. Nature. 2011;474(7353):609-615. doi:10.1038/nature10166
[4] Konstantinopoulos PA, Matulonis UA. Clinical and translational advances in ovarian cancer therapy. Nat Cancer. 2023;4(9):1239-1257. doi:10.1038/s43018-023-00617-9
[5] Wu L, Yang S, Li H, Zhang Y, Feng L, Zhang C, Wei J, Gu X, Xu G, Wang Z, Wang F. TSPAN4-positive migrasome derived from retinal pigmented epithelium cells contributes to the development of proliferative vitreoretinopathy. J Nanobiotechnology. 2022;20(1):519. doi:10.1186/s12951-022-01732-y
[6] Ma L, Li Y, Peng J, Wu D, Zhao X, Cui Y, Chen L, Yan X, Du Y, Yu L. Discovery of the migrasome, an organelle mediating release of cytoplasmic contents during cell migration. Cell Res. 2015;25(1):24-38. doi:10.1038/cr.2014.135
[7] Strzyz P. Migrasomes promote angiogenesis. Nat Rev Mol Cell Biol. 2023;24(2):84. doi:10.1038/s41580-022-00570-y
[8] Qin Y, Yang J, Liang C, Liu J, Deng Z, Yan B, Fu Y, Luo Y, Li X, Wei X, Li W. Pan-cancer analysis identifies migrasome-related genes as a potential immunotherapeutic target: A bulk omics research and single cell sequencing validation. Front Immunol. 2022;13:994828. doi:10.3389/fimmu.2022.994828
[9] Derrien T, Johnson R, Bussotti G, Tanzer A, Djebali S, Tilgner H, Guernec G, Martin D, Merkel A, Knowles DG, Lagarde J, Veeravalli L, Ruan X, Ruan Y, Lassmann T, Carninci P, Brown JB, Lipovich L, Gonzalez JM, Thomas M, Davis CA, Shiekhattar R, Gingeras TR, Hubbard TJ, Notredame C, Harrow J, Guigó R. The GENCODE v7 catalog of human long noncoding RNAs: analysis of their gene structure, evolution, and expression. Genome Res. 2012;22(9):1775-1789. doi:10.1101/gr.132159.111
[10] Mukherjee N, Calviello L, Hirsekorn A, de Pretis S, Pelizzola M, Ohler U. Integrative classification of human coding and noncoding genes through RNA metabolism profiles. Nat Struct Mol Biol. 2017;24(1):86-96. doi:10.1038/nsmb.3325
[11] Liu SJ, Horlbeck MA, Cho SW, Birk HS, Malatesta M, He D, Attenello FJ, Villalta JE, Cho MY, Chen Y, Mandegar MA, Olvera MP, Gilbert LA, Conklin BR, Chang HY, Weissman JS, Lim DA. CRISPRi-based genome-scale identification of functional long noncoding RNA loci in human cells. Science. 2017;355(6320):aah7111. doi:10.1126/science.aah7111
[12] Cooper LA, Demicco EG, Saltz JH, Powell RT, Rao A, Lazar AJ. PanCancer insights from The Cancer Genome Atlas: the pathologist’s perspective. J Pathol. 2018;244(5):512-524. doi:10.1002/path.5028
[13] Zhao X, Lei Y, Zheng J, Peng J, Li Y, Yu L, Chen Y. Identification of markers for migrasome detection. Cell Discov. 2019;5:27. doi:10.1038/s41421-019-0093-y
[14] Zheng Y, Lang Y, Qi B, Li T. TSPAN4 and migrasomes in atherosclerosis regression correlated to myocardial infarction and pan-cancer progression. Cell Adh Migr. 2023;17(1):14-19. doi:10.1080/19336918.2022.2155337
[15] Stelzer G, Rosen N, Plaschkes I, Zimmerman S, Twik M, Fishilevich S, Stein TI, Nudel R, Lieder I, Mazor Y, Kaplan S, Dahary D, Warshawsky D, Guan-Golan Y, Kohn A, Rappaport N, Safran M, Lancet D. The GeneCards Suite: From Gene Data Mining to Disease Genome Sequence Analyses. Curr Protoc Bioinformatics. 2016;54:1.30.1-1.30.33. doi:10.1002/cpbi.5
[16] Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, Smyth GK. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 2015;43(7):e47. doi:10.1093/nar/gkv007
[17] Cai W, van der Laan M. Nonparametric bootstrap inference for the targeted highly adaptive least absolute shrinkage and selection operator (LASSO) estimator. Int J Biostat. 2020. doi:10.1515/ijb-2017-0070
[18] Engebretsen S, Bohlin J. Statistical predictions with glmnet. Clin Epigenetics. 2019;11(1):123. doi:10.1186/s13148-019-0730-1
[19] Yu G, Wang LG, Han Y, He QY. clusterProfiler: an R package for comparing biological themes among gene clusters. OMICS. 2012;16(5):284-287. doi:10.1089/omi.2011.0118
[20] Hänzelmann S, Castelo R, Guinney J. GSVA: gene set variation analysis for microarray and RNA-seq data. BMC Bioinformatics. 2013;14:7. doi:10.1186/1471-2105-14-7
[21] Roett MA, Evans P. Ovarian cancer: an overview. Am Fam Physician. 2009;80(6):609-616. PMID:19817326
[22] Deng S, Wu Y, Huang S, Yang X. Novel insights into the roles of migrasome in cancer. Discov Oncol. 2024;15(1):166. doi:10.1007/s12672-024-00942-0
[23] Wang LJ, Xu R, Wu Y. Migrasome regulator TSPAN4 shapes the suppressive tumor immune microenvironment in pan-cancer. Front Immunol. 2024;15:1419420. doi:10.3389/fimmu.2024.1419420
[24] Chen J, Yang L, Ma Y, Zhang Y. Recent advances in understanding the immune microenvironment in ovarian cancer. Front Immunol. 2024;15:1412328. doi:10.3389/fimmu.2024.1412328
[25] Hu Z, Yuan L, Yang X, Yi C, Lu J. The roles of long non-coding RNAs in ovarian cancer: from functions to therapeutic implications. Front Oncol. 2024;14:1332528. doi:10.3389/fonc.2024.1332528
[26] Yoshihara K, Shahmoradgoli M, Martínez E, Vegesna R, Kim H, Torres-Garcia W, Treviño V, Shen H, Laird PW, Levine DA, Carter SL, Getz G, Stemke-Hale K, Mills GB, Verhaak RGW. Inferring tumour purity and stromal and immune cell admixture from expression data. Nat Commun. 2013;4:2612. doi:10.1038/ncomms3612
[27] Newman AM, Liu CL, Green MR, Gentles AJ, Feng W, Xu Y, Hoang CD, Diehn M, Alizadeh AA. Robust enumeration of cell subsets from tissue expression profiles. Nat Methods. 2015;12(5):453-457. doi:10.1038/nmeth.3337
[28] Mayakonda A, Lin DC, Assenov Y, Plass C, Koeffler HP. Maftools: efficient and comprehensive analysis of somatic variants in cancer. Genome Res. 2018;28(11):1747-1756. doi:10.1101/gr.239244.118
[29] Jiang P, Gu S, Pan D, Fu J, Sahu A, Hu X, Li Z, Traugh N, Bu X, Li B, Liu J, Freeman GJ, Brown MA, Wucherpfennig KW, Liu XS. Signatures of T cell dysfunction and exclusion predict cancer immunotherapy response. Nat Med. 2018;24(10):1550-1558. doi:10.1038/s41591-018-0136-1
[30] Maeser D, Gruener RF, Huang RS. oncoPredict: an R package for predicting in vivo or cancer patient drug response and biomarkers from cell line screening data. Brief Bioinform. 2021;22(6):bbab260. doi:10.1093/bib/bbab260
[31] Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, Paulovich A, Pomeroy SL, Golub TR, Lander ES, Mesirov JP. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A. 2005;102(43):15545-15550. doi:10.1073/pnas.0506580102
[32] Tibshirani R. Regression shrinkage and selection via the lasso. J R Stat Soc Series B Stat Methodol. 1996;58(1):267-288. doi:10.1111/j.2517-6161.1996.tb02080.x
[33] Friedman J, Hastie T, Tibshirani R. Regularization paths for generalized linear models via coordinate descent. J Stat Softw. 2010;33(1):1-22. doi:10.18637/jss.v033.i01
[34] Yang W, Soares J, Greninger P, Edelman EJ, Lightfoot H, Forbes S, Bindal N, Beare D, Smith JA, Thompson IR, Ramaswamy S, Futreal PA, Haber DA, Stratton MR, Benes C, McDermott U, Garnett MJ. Genomics of Drug Sensitivity in Cancer (GDSC): a resource for therapeutic biomarker discovery in cancer cells. Nucleic Acids Res. 2013;41(Database issue):D955-D961. doi:10.1093/nar/gks1111
[35] Shang A, Wang W, Gu C, Chen C, Zeng B, Yang Y, Ji P, Sun J, Wu J, Lu W, Sun Z. Long non-coding RNA HOTTIP enhances IL-6 expression to potentiate immune escape of ovarian cancer cells by upregulating the expression of PD-L1 in neutrophils. J Exp Clin Cancer Res. 2019;38(1):411. doi:10.1186/s13046-019-1394-6
- Qian M, Ling W, Ruan Z. Long non-coding RNA SNHG12 promotes immune escape of ovarian cancer cells through their crosstalk with M2 macrophages. Aging (Albany NY). 2020;12(17):17122-17136. doi:10.18632/aging.103653