Li Y, Nie Y, Guo H, Guo H, Ha C, Li Y. Establish of an Initial Platinum-Resistance Predictor in High-Grade Serous Ovarian Cancer Patients Regardless of Homologous Recombination Deficiency Status.
Front Oncol 2022;
12:847085. [PMID:
35372049 PMCID:
PMC8971787 DOI:
10.3389/fonc.2022.847085]
[Citation(s) in RCA: 2] [Impact Index Per Article: 1.0] [Reference Citation Analysis] [Abstract] [Grants] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 01/01/2022] [Accepted: 02/25/2022] [Indexed: 11/13/2022] Open
Abstract
Backgrounds
Ovarian cancer (OC) is still the leading aggressive and lethal disease of gynecological cancers, and platinum-based regimes are the standard treatments. However, nearly 20%–30% of patients with OC are initial platinum resistant (IPR), and there is a lack of valid tools to predict whether they will be primary platinum resistant or not prior to chemotherapy.
Methods
Transcriptome data from The Cancer Genome Atlas (TCGA) was downloaded as the training data, and transcriptome data of GSE15622, GSE102073, GSE19829, and GSE26712 were retrieved from Gene Expression Omnibus (GEO) as the validation cohorts. Differentially expressed genes (DEGs) were selected between platinum-sensitive and platinum-resistant patients from the training cohort, and multiple machine-learning algorithms [including random forest, XGboost, and least absolute shrinkage and selection operator (LASSO) regression] were utilized to determine the candidate genes from DEGs. Then, we applied logistic regression to establish the IPR signature based on the expression. Finally, comprehensive clinical, genomic, and survival feature were analyzed to understand the application value of the established IPR signature.
Results
A total of 532 DEGs were identified between platinum-resistant and platinum-sensitive samples, and 11 of them were shared by these three-machine learning algorithms and utilized to construct an IPR prediction signature. The area under receiver operating characteristic curve (AUC) was 0.841 and 0.796 in the training and validation cohorts, respectively. Notably, the prediction capacity of this signature was stable and robust regardless of the patients’ homologous recombination deficiency (HRD) and mutation burden status. Meanwhile, the genomic feature was concordant between samples with high- or low-IPR signature, except a significantly higher prevalence of gain at Chr19q.12 (regions including CCNE1) in the high-IPR signature samples. The efficacy of prediction of platinum resistance of IPR signature successfully transferred to the precise survival prediction, with the AUC of 0.71, 0.72, and 0.66 to predict 1-, 3-, and 5-year survival, respectively. At last, we found a significantly different tumor-infiltrated lymphocytes feature, including lower abundance of CD4+ naive T cells in the samples with high-IPR signature. A relatively lower tumor immune dysfunction and exclusion (TIDE) value and more sensitivity to multiple therapies including Gefitinib may suggest the potency to transfer from platinum-based therapy to immunotherapy or target therapies in patients with high-IPR signature.
Conclusion
Our study established an IPR signature based on the expression of 11 genes that could stably and robustly distinguish OC patients with IPR and/or poor outcomes, which may guide therapeutic regimes tailoring.
Collapse