Related Experiment Video
Updated: Sep 25, 2025
MicroRNA Based Liquid Biopsy: The Experience of the Plasma miRNA Signature Classifier MSC for Lung Cancer Screening
Published on: October 26, 2017
Construction of a miRNA Signature Using Support Vector Machine to Identify Microsatellite Instability Status and
Ya-Nan Wang1, Ya-Ning Wei2, Xiang-Yu Fan1
1Department of Pathology, Affiliated Hospital of Hebei University, Baoding, China.
Background:
The specific role and prognostic value of DNA repair and replication-associated miRNAs in gastric cancer (GC) have not been clearly elucidated. Therefore, comprehensive analysis of miRNAs in GC is crucial for proposing therapeutic strategies and survival prediction.
Methods:
Firstly, clinical information and transcriptome data of TCGA-GC were downloaded from the database. In the entire cohort, we performed differential analysis in all miRNAs and support vector machine (SVM) was used to eliminate redundant miRNAs. Subsequently, we combined survival data and cox regression analysis to construct a miRNA signature in the training cohort. In addition, we used PCA, Kaplan-Meier, and ROC analysis to explore the prognosis value of risk score in the training and testing cohort. It is worth noting that multiple algorithms were used to evaluate difference of immune microenvironment (TME), microsatellite instability (MSI), tumor mutational burden (TMB), and immunotherapy in different risk groups. Finally, we investigated the potential mechanism about miRNA signature.
Results:
We constructed miRNA signature based on the following 4 miRNAs: hsa-miR-139-5p, hsa-miR-139-3p, hsa-miR-146b-5p, and hsa-miR-181a-3p. Univariate and multivariate Cox regression analyses suggested that risk score is a risk factor and an independent prognostic factor in GC patients. The AUC value of ROC analysis showed a robust prediction accuracy in each cohort. Moreover, significant differences in immune functions, immune cell content, immune checkpoint, MSI status, and TMB score were excavated in different groups distinguished by risk score. Finally, based on the above four miRNA target genes, we revealed that the signature was enriched in DNA repair and replication.
Conclusion:
We have developed a robust risk-formula based on 4 miRNAs that provides accurate risk stratification and prognostic prediction for GC patients. In addition, different risk subgroups may potentially guide the choice of targeted therapy.
