Discovery of key molecular signatures for diagnosis and therapies of glioblastoma by combining supervised and unsupervised learning approaches

Abstract

Glioblastoma (GBM) is the most malignant brain cancer and one of the leading causes of cancer-related death globally. So, identifying potential molecular signatures and associated drug molecules are crucial for diagnosis and therapies of GBM. This study suggested GBM-causing ten key genes (ASPM, CCNB2, CDK1, AURKA, TOP2A, CHEK1, CDCA8, SMC4, MCM10, and RAD51AP1) from nine transcriptomics datasets by combining supervised and unsupervised learning results. Differential expression patterns of key genes (KGs) between GBM and control samples were verified by different independent databases. Gene regulatory network (GRN) detected some important transcriptional and post-transcriptional regulators for KGs. The KGs-set enrichment analysis unveiled some crucial GBM-causing molecular functions, biological processes, cellular components, and pathways. The DNA methylation analysis detected some hypo-methylated CpG sites that might stimulate the GBM development. From the immune infiltration analysis, we found that almost all KGs are associated with different immune cell infiltration levels. Finally, we recommended KGs-guided four repurposable drug molecules (Fluoxetine, Vatalanib, TGX221 and RO3306) against GBM through molecular docking, drug likeness, ADMET analyses and molecular dynamics simulation studies. Thus, the discoveries of this study could serve as valuable resources for wet-lab experiments in order to take a proper treatment plan against GBM.

Keywords: Bioinformatics and machine learning approaches; Drug repurposing; Gene expression profiles; Glioblastoma; Key genes.

Fig. 1

The workflow of the study.

Fig. 2

Cluster dendrogram of merged and unmerged modules for (A) GSE68848 (B) GSE86574 and (C) GSE104292.

Fig. 3

The volcano plot based on ACC and LogFC. (A) DEGs selection by RF-based prediction model (B) DEGs selection by SVM-based prediction model.

Fig. 4

The cDEGs between the ML and WGCNA techniques are depicted in a Venn diagram.

Fig. 5

PPI network study of cDEGs. Green nodes indicate lower interactions, yellow nodes indicate medium interactions and red nodes indicate higher interactions (KGs).

Fig. 6

Binding affinities between receptor proteins and drug agents.

Fig. 7

100ns MD simulation results for the top four complexes. (A) The RMSDs’ periodic progression. (B) MM-PBSA binding free energy (kJ/mol).