A Novel Two-Gene Expression-Based Prognostic Score in Malignant Pleural Mesothelioma

Our plain-English summary of A Novel Two-Gene Expression-Based Prognostic Score in Malignant Pleural Mesothelioma, by V. Shivarov, G. Blazhev, A. Yordanov (Diagnostics, 2023).

The study

• Authors: V. Shivarov, G. Blazhev, A. Yordanov
• Journal: Diagnostics
• Published: 2023;13(9)
• Cited by: 1 article, as counted by NIH iCite
• Read it: PubMed · Publisher · Free full text

Abstract

Background: Malignant pleural mesothelioma (MPM) is a rare cancer type with an increasing incidence worldwide. We aimed to develop a rational gene expression-based prognostic score in MPM using publicly available datasets.

Methods: We developed and validated a two-gene prognostic score (2-PS) using three independent publicly available gene expression datasets. The 2-PS was built using the Robust Likelihood-Based Survival Modeling with Microarray Data method.

Results: We narrowed down the model building to the analysis of 179 genes, which have been shown previously to be of importance to MPM development. Our statistical approach showed that a model including two genes (GOLT1B and MAD2L1) was the best predictor for overall survival (OS) (p < 0.0001). The binary score based on the median of the continuous score stratified the patients into low and high score groups and also showed statistical significance in uni- and multivariate models. The 2-PS was validated using two independent transcriptomic datasets. Furthermore, gene set enrichment analysis using training and validation datasets showed that high score patients had distinct gene expression profiles. Our 2-PS also showed a correlation with the estimated infiltration by some immune cell fractions such as CD8+ T cells and M1/2 macrophages. Finally, 2-PS correlated with sensitivity or resistance to some commonly used chemotherapeutic drugs.

Conclusion: This is the first study to demonstrate good performance of only two-gene expression-based prognostic scores in MPM. Our initial approach for features selection allowed for an increased likelihood for the predictive value of the developed score, which we were also able to demonstrate.

In plain English

This study did not examine asbestos exposure. Using publicly available gene expression data (measurements of how active genes are) from tumours of people with malignant pleural mesothelioma, the authors built a score meant to predict overall survival. They started from 179 genes that mesothelioma cell lines had been reported to depend on. A statistical method then selected a two-gene model (GOLT1B and MAD2L1) using data on 87 cases from The Cancer Genome Atlas. They tested the score in two other published datasets, of 211 cases (Bueno) and 67 cases (Blum). In each dataset, patients were split into high and low scores at that dataset's median. The score was statistically significantly associated with overall survival, and it stayed significant in models that included other clinical factors. Its area under the ROC curve was 0.67, 0.75 and 0.85 in the three datasets. The score also correlated with immune cell estimates and, in laboratory cell lines, with response to some drugs, including cisplatin. The authors called its prognostic power moderate. They said the dataset-specific cut-offs cannot be applied directly and that prospective validation is needed.

We wrote this plain-English summary from the full article with the help of AI, and a second AI model checked each sentence against the article. It is not the authors' wording; the abstract and the article are their own account.

What the study did and found

• Design: Bioinformatic analysis of publicly available tumour gene expression datasets: development and validation of a gene expression-based prognostic score, plus analyses of mesothelioma cell line drug sensitivity data
• Who was studied: Malignant pleural mesothelioma (MPM) cases in public datasets: 87 cases from The Cancer Genome Atlas (TCGA) for training; 211 cases (Bueno, RNA sequencing) and 67 cases (Blum, expression arrays) for validation; described as two cohorts from the USA and one from Europe. Drug analyses used 16 and 15 MPM cell lines. Years of patient collection are not given.
• Exposure: Asbestos exposure was not studied. The analysis concerned tumour expression of two genes (GOLT1B and MAD2L1) combined into a score. Asbestos is mentioned only as background in the introduction.
• Main results:
  • Starting from 179 genes, the statistical method selected a best-performing survival model of two genes, GOLT1B and MAD2L1, in the TCGA training data
  • TCGA (training) cohort: the median-split binary score had an area under the ROC curve of 0.67
  • TCGA cohort: in a multivariate model with age, sex, stage, histology and mutational status, the binary score kept independent prognostic value (Cox regression p = 7.34 times 10 to the minus 6)
  • Bueno validation cohort (n = 211): the binary score had prognostic power (Cox regression p = 1.4 times 10 to the minus 5) with an area under the ROC curve of 0.75
  • Bueno validation cohort: in a multivariate model the binary score was still of independent prognostic value (Cox regression p = 0.03)
  • Blum validation cohort (n = 67): the binary score had an area under the ROC curve of 0.85 (Cox regression p = 0.0014)
  • Blum cohort: in a multivariate model with sex, age, stage and histology (no mutational data), the binary score kept independent prognostic value (Cox regression p = 0.00671)
  • Gene set analysis: 25 signatures were enriched in high-score patients in all three cohorts. Most were related to DNA repair and DNA damage response.
  • In both the TCGA and Bueno datasets, the continuous score correlated positively with estimated CD8+ T cell fraction and with M1 and M2 macrophage fractions
  • MPM cell lines: the score correlated with response to 11 drugs in the GDSC1 set (16 cell lines) and 18 drugs in the GDSC2 set (15 cell lines). It correlated inversely with the cisplatin dose-response area under the curve (R = −0.51, p = 0.046), which the authors read as possibly predicting higher sensitivity to cisplatin.
  • MPM cell lines: the score correlated positively with the dose-response area under the curve for gemcitabine (R = 0.69, p = 0.019) and vinblastine (R = 0.63, p = 0.037), which the authors read as suggesting resistance to those drugs
• The authors' conclusion: The authors concluded that they had developed a two-gene expression-based prognostic score in MPM and validated it in two independent cohorts. They said it defines patient subgroups with specific gene expression, immune features and drug sensitivity, and that it can be tested prospectively with methods such as RT-PCR and immunohistochemistry.
• Limitations the authors state:
  • The authors described the score's prognostic power as moderate
  • The high/low cut-offs were each dataset's own median, so they cannot be applied directly to prospective studies
  • The model still needs prospective validation with a simple, standardized measurement technique
  • Mutational data were not available for the Blum multivariate model
  • The drug-sensitivity analysis in cell lines was described as preliminary evidence of predictive power
  • The score's performance may depend on how patients in each cohort were treated, so further evaluation should focus on uniformly treated patients

Topics

Medical terms in this study, explained in our Lexicon: Pleural Mesothelioma, Overall Survival, Statistical Significance, Macrophage.

Related studies

• Identification of a novel prognostic gene signature in pleural mesothelioma: a study based on The Cancer Genome Atlas database and experimental validation — X. Wang, Y. Yang, et al. (2025)
• Differential expression of DMD transcripts as a novel prognostic biomarker in histologically diverse mesotheliomas — N. Alnassar, J.M.J. Derry, et al. (2024)
• SLC17A9 expression levels in a pan‑cancer panel and validation of the role of SLC17A9 as a novel prognostic biomarker for osteosarcoma — J. Li, F. Wu, et al. (2023)
• MUC4, a novel immunohistochemical marker identified by gene expression profiling, differentiates pleural sarcomatoid mesothelioma from lung sarcomatoid carcinoma — V.J. Amatya, K. Kushitani, et al. (2017)
• Assessment of Glutathione Peroxidase-1 (GPX1) Gene Expression as a Specific Diagnostic and Prognostic Biomarker in Malignant Pleural Mesothelioma — A.F. Gharib, M. Alaa Eldeen, et al. (2021)
• Gene Expression Analysis of Biphasic Pleural Mesothelioma: New Potential Diagnostic and Prognostic Markers — R. Bruno, A.M. Poma, et al. (2022)

Browse more research in Diagnostics, or return to medical research.