Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2024 Aug 24;10(17):e36898.
doi: 10.1016/j.heliyon.2024.e36898. eCollection 2024 Sep 15.

A clinical prognostic model related to T cells based on machine learning for predicting the prognosis and immune response of ovarian cancer

Affiliations

A clinical prognostic model related to T cells based on machine learning for predicting the prognosis and immune response of ovarian cancer

Qiwang Lin et al. Heliyon. .

Abstract

Background: Ovarian cancer (OV) is regarded as one of the most lethal malignancies affecting the female reproductive system, with individuals diagnosed with OV often facing a dismal prognosis due to resistance to chemotherapy and the presence of an immunosuppressive environment. T cells serve as a crucial mediator for immune surveillance and cancer elimination. This study aims to analyze the mechanism of T cell-associated markers in OV and create a prognostic model for clinical use in enhancing outcomes for OV patients.

Methods: Based on the single-cell dataset GSE184880, this study used single-cell data analysis to identify characteristic T cell subsets. Analysis of high dimensional weighted gene co-expression network analysis (hdWGCNA) is utilized to identify crucial gene modules along with their corresponding hub genes. A grand total of 113 predictive models were formed utilizing ten distinct machine learning algorithms along with the combination of the cancer genome atlas (TCGA)-OV dataset and the GSE140082 dataset. The most dependable clinical prognostic model was created utilizing the leave one out cross validation (LOOCV) framework. The validation process for the models was achieved by conducting survival curve analysis and receiver operating characteristic (ROC) analysis. The relationship between risk scores and immune cells was explored through the utilization of the Cibersort algorithm. Additionally, an analysis of drug sensitivity was carried out to anticipate chemotherapy responses across various risk groups. The genes implicated in the model were authenticated utilizing qRT-PCR, cell viability experiments, and EdU assay.

Results: This study developed a clinical prognostic model that includes ten risk genes. The results obtained from the training set of the study indicate that patients classified in the low-risk group experience a significant survival advantage compared to those in the high-risk group. The ROC analysis demonstrates that the model holds significant clinical utility. These results were verified using an independent dataset, strengthening the model's precision and dependability. The risk assessment provided by the model also serves as an independent prognostic factor for OV patients. The study also unveiled a noteworthy relationship between the risk scores calculated by the model and various immune cells, suggesting that the model may potentially serve as a valuable tool in forecasting responses to both immune therapy and chemotherapy in ovarian cancer patients. Notably, experimental evidence suggests that PFN1, one of the genes included in the model, is upregulated in human OV cell lines and has the capacity to promote cancer progression in in vitro models.

Conclusion: We have created an accurate and dependable clinical prognostic model for OV capable of predicting clinical outcomes and categorizing patients. This model effectively forecasts responses to both immune therapy and chemotherapy. By regulating the immune microenvironment and targeting the key gene PFN1, it may improve the prognosis for high-risk patients.

Keywords: Clinical prognostic model; Machine learning; Ovarian cancer; PFN1; Single-cell analysis.

PubMed Disclaimer

Conflict of interest statement

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Figures

Fig. 1
Fig. 1
Single cell analysis identified key T cell subsets. (A) Through UMAP analysis, a total of 15 cell clusters were identified in all samples. (B) 15 cell clusters are annotated as 8 highly heterogeneous cell types. (C) The proportion of 8 cell types between the OV sample group and the normal sample group. (D) The T cell subpopulations were subjected to another dimensionality reduction clustering analysis and displayed through the UMAP graph. (E) The proportion of T cell subpopulation related clusters between the OV sample group and the normal sample group.
Fig. 2
Fig. 2
HdWGCNA shows that turquoise, pink, yellow, red, blue, black, and green modules are seven central modules closely related to T cell subsets in OV. (A) Choose soft power = 6 to build a scale-free network. (B) Employ a tree diagram to depict the nine modules within a scale-free network. (C) Identified eight gene modules and showcased the principal genes via the hdWGCNA pipeline. (D) The eight gene modules' distribution across various T cell subpopulations. (E) Conduct correlation analysis among the different gene modules. (F) Explain the bubble diagram correlating the module with T cell subpopulations.
Fig. 3
Fig. 3
Functional annotation analysis of hub genes. (A) Gene Ontology analysis of hub genes. (B) Kyoto Encyclopedia of Genes and Genomes analysis of hub genes. (C–E) Mapping a trajectory of differentiation based on cell differentiation status, coloring cell development time, and identifying cell clusters. (F) The expression heatmap of genes with simulated time values shows that genes with similar expression trends converge to form different clusters.
Fig. 4
Fig. 4
Construct a clinical prognosis model and evaluate it using machine learning algorithms. (A, B) Differential expression analysis of hub genes between normal and OV samples. A: Heat map; B: Volcano map. (C) Calculate the C-index for each prognostic model created using 10 machine learning algorithms and 113 combinations in both the training and validation datasets. (D) Examine the frequency of copy number variations (CNVs) in 10 selected genes in TCGA-OV. Green indicates deletion frequency, while red dots represent amplification frequency. (E) Show the genomic locations of the 10 model genes on chromosomes. (F, G) Evaluate the survival and receiver operating characteristic (ROC) curves for OV patients classified into high-risk and low-risk groups in the training and validation datasets. F: TCGA-OV; G: GSE140082. (H) Perform univariate and multivariate Cox regression analyses considering clinical characteristics and risk scores of the clinical prognostic models in the training and validation datasets.
Fig. 5
Fig. 5
Correlation between the risk score of clinical prognostic models and the TME of OV. (A) The proportion of immune cell infiltration is shown for each sample. (B) A comparison of immune cell infiltration levels is presented between high-risk and low-risk groups. (C) Displays the relationship between risk scores and immune cell presence. (D) Shows the correlation between risk scores and various clinical features. (E–F) Waterfall plots highlight somatic mutations in populations separated by risk level (E for high-risk; F for low-risk groups).
Fig. 6
Fig. 6
Drug sensitivity analysis of high-risk and low-risk populations. (A) AZD1332. (B) BMS -536924. (C) BMS -754807. (D) ERK_2440. (E) Foretinib. (F) NVP-ADW742. (G) Taselisib. (H)ML323.
Fig. 7
Fig. 7
PFN1 is upregulated in human OV cell line A2780. (A–J) PCR was conducted to assess the transcript levels of TGOLN2, NBL1, ARID1B, ISG20, KRAS, CLIC3, ALOX5AP, PFN1, DNAJA1, and LASP1 in human normal ovarian epithelial cells (IOSE80) and the human OV cell line A2780. N = 3/group. ***≤0.001.
Fig. 8
Fig. 8
PFN1 is upregulated in human OV cell line HEY. (A–J) PCR was performed to detect the transcript levels of TGOLN2, NBL1, ARID1B, ISG20, KRAS, CLIC3, ALOX5AP, PFN1, DNAJA1, and LASP1 in human normal ovarian epithelial cells, IOSE80, and in human OV cell line, HEY. N = 3/group. **≤0.01.
Fig. 9
Fig. 9
PFN1 promotes the progression of OV in vitro. (A) Detection of the efficiency of PFN1 expression inhibited by small interfering RNA in the A2780 cell line. (B) Changes in cell line viability of A2780 cells after inhibition of PFN1. (C) Detection of the efficiency of PFN1 expression inhibited by small interfering RNA in the HEY cell line. (D) Changes in cell line viability of HEY cells after inhibition of PFN1. (E–G) Alterations in the proliferative capacity of A2780 and HEY cell lines after inhibition of PFN1. (H–K) Alterations in the transcription of P21, BAX, CDH1 and CDH2 following inhibition of PFN1 in the A2780 cell line. (L–O) Alterations in the transcription of P21, BAX, CDH1 and CDH2 following inhibition of PFN1 in the HEY cell line. N = 3. *≤0.05, **≤0.01, ***≤0.001, ****≤0.0001. The results are presented as mean ± SD.

Similar articles

Cited by

References

    1. Konstantinopoulos P.A., Matulonis U.A. Clinical and translational advances in ovarian cancer therapy. Nature cancer. 2023;4(9):1239–1257. - PubMed
    1. Jacobs I.J., Menon U., Ryan A., et al. Ovarian cancer screening and mortality in the UK Collaborative Trial of Ovarian Cancer Screening (UKCTOCS): a randomised controlled trial. Lancet (London, England) 2016;387(10022):945–956. - PMC - PubMed
    1. Ebell M.H., Culp M.B., Radke T.J. A systematic review of symptoms for the diagnosis of ovarian cancer. Am. J. Prev. Med. 2016;50(3):384–394. - PubMed
    1. Bray F., Ferlay J., Soerjomataram I., Siegel R.L., Torre L.A., Jemal A. Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA A Cancer J. Clin. 2018;68(6):394–424. - PubMed
    1. Guo Y., Li S., Li C., Wang L., Ning W. Multifactor assessment of ovarian cancer reveals immunologically interpretable molecular subtypes with distinct prognoses. Front. Immunol. 2023;14 - PMC - PubMed

LinkOut - more resources