DeMoS: dense module based gene signature detection through quasi-clique: an application to cervical cancer prognosis
Article Type
Research Article
Publication Title
Network Modeling Analysis in Health Informatics and Bioinformatics
Abstract
Nowadays, cervical cancer is a leading cause of death among women. Determining the gene signature is one of the major issues in bioinformatics. Though many of the methodologies and applications have been given as suggestions in recent literature, efficient techniques, which may be considered complex gene expression profiles, will be able to find out the most relevant signatures required. In the given article, we demonstrate a new framework to find out the dense module-based gene signatures (DeMoS) and their targeting miRNAs using the quasi-clique detection algorithm and discuss their application in the field of prognostic survival studies. We used a cervical cancer data repository with prognostic clinical data to conduct this experiment. At first, we executed the empirical Bayes test by applying the linear model for the microarray method to find out the dysregulated genes, or miRNAs. MiRNA-mediated dysregulated target genes were pulled out of the particular dysregulated miRNAs. After that, we discovered densely co-expressed modules by applying a quasi-clique identification technique. The average correlation coefficient has been computed for each resultant module, and the module that contains the highest correlation was composed as the resultant gene signature (10-gene signatures containing ten genes are as follows: FGF9, FGF18, PPP1R9A, ERBB4, DCDC2, TOX3, ARMC3, DNALI1, RGL3, and ENPP3). After that, we applied 10-fold cross-validation to three common classifiers (SVM, PAM, and random forest) and obtained the AUC. (0.95 for SVM, 0.955 for RF, and 0.955 for PAM) that is better than the state-of-the-art algorithms (Li et al. in Technol Cancer Res Treat 17:1533033818767455, 2017/2018; Huang et al. in Cancer 117(15):3363–3373, 2011). In addition to it, we found eight dysregulated miRNAs that have targeted the gene, as mentioned earlier. At last, we performed a prognosis survival study for the resultant gene signature (i.e., containing the p-value of Cox regression as 4.2e-02). DEMOS is very useful for determining the signature for any of the microarray or RNA-Seq profiles. The code is available at https://github.com/sahasuparna/DeMoS.
DOI
10.1007/s13721-024-00470-5
Publication Date
12-1-2024
Recommended Citation
Saha, Suparna; Seth, Soumita; Ghosh, Soumadip; Qin, Guimin; Bhadra, Tapas; Pati, Soumen Kumar; Chakraborty, Somenath; and Mallik, Saurav, "DeMoS: dense module based gene signature detection through quasi-clique: an application to cervical cancer prognosis" (2024). Journal Articles. 4701.
https://digitalcommons.isical.ac.in/journal-articles/4701
Comments
Open Access; Green Open Access