LASSO-based screening for potential prognostic biomarkers associated with glioblastoma.

BACKGROUND

Glioblastoma is the most common malignancy of the neuroepithelium, yet existing research on this tumor is limited. LASSO is an algorithm of selected feature coefficients by which genes associated with glioblastoma prognosis can be obtained.

METHODS

Glioblastoma-related data were selected from the Cancer Genome Atlas (TCGA) database, and information was obtained for 158 samples, including 153 cancer samples and five samples of paracancerous tissue. In addition, 2,642 normal samples were selected from the Genotype-Tissue Expression (GTEx) database. Whole-gene bulk survival analysis and differential expression analysis were performed on glioblastoma genes, and their intersections were taken. Finally, we determined which genes are associated with glioma prognosis. The STRING database was used to analyze the interaction network between genes, and the MCODE plugin under Cytoscape was used to identify the highest-scoring clusters. LASSO prognostic analysis was performed to identify the key genes. Gene expression validation allowed us to obtain genes with significant expression differences in glioblastoma cancer samples and paracancer samples, and glioblastoma independent prognostic factors could be derived by univariate and multivariate Cox analyses. GO functional enrichment analysis was performed, and the expression of the screened genes was detected using qRT-PCR.

RESULTS

Whole-gene bulk survival analysis of glioblastoma genes yielded 607 genes associated with glioblastoma prognosis, differential expression analysis yielded 8,801 genes, and the intersection of prognostic genes with differentially expressed genes (DEG) yielded 323 intersecting genes. PPI analysis of the intersecting genes revealed that the genes were significantly enriched in functions such as the formation of a pool of free 40S subunits and placenta development, and the highest-scoring clusters were obtained using the MCODE plug-in. Eight genes associated with glioblastoma prognosis were identified based on LASSO analysis: RPS10, RPS11, RPS19, RSL24D1, RPL39L, EIF3E, NUDT5, and RPF1. All eight genes were found to be highly expressed in the tumor by gene expression verification, and univariate and multivariate Cox analyses were performed on these eight genes to identify RPL39L and NUDT5 as two independent prognostic factors associated with glioblastoma. Both RPL39L and NUDT5 were highly expressed in glioblastoma cells.

CONCLUSION

Two independent prognostic factors in glioblastoma, RPL39L and NUDT5, were identified.