Improved identification of core biomarkers and drug repositioning for ovarian cancer: an integrated bioinformatics approach

Abstract

Ovarian cancer (OC) is a frequent fatal malignancy in female reproductive systems with poor early diagnosis. There are several currently utilized popular methods for the detection of OC biomarkers from the microarray transcriptomics dataset, but these methods suffer from the limitation in the identification of biomarkers in the presence of outlier. In this study, we introduced a rule for modification of outlier to improve the performance of biomarker selection methods. We employed the proposed procedure on simulated and three publicly available ovarian cancer gene expression datasets, and improved performance of the proposed procedure was observed. We identified 226 differentially expressed genes (DEGs) overlapped in 3 proposed modified microarray OC datasets using LIMMA in R. These DEGs were underwent Gene Ontology analysis and revealed apoptotic signaling and programmed cell death as an important biological process. The pathway enrichment analysis showed molecular pathways in OC. We also identified FOXC1, GATA2, E2F1, YY1, and FOXL1 as regulatory transcription factors. The protein–protein interaction analysis demonstrated upregulated hub proteins (HDAC1, RPS15, SF3B1, YWHAH, EIF1AX, CALM1, PSME3, UBC, MCL1, NFE2L2), and down-regulated hub proteins (RPL7, RPL9, HSP90AB1, RPS16, RPL30, SKP1, RPL10, RPL14, RPL24, RPLP1). The differential expression of these hub proteins was cross-validated in independent OC RNA-Seq datasets from the TCGA database. The prognostic performance of these hub proteins was observed associated with the worst survival outcomes in OC. Finally, considering 226 DEGs as gene signature, using the L1000CDS2, we revealed 18 drugs in OC with overlap > 0.04. The predicted drugs were repositioned in OC considering the core biomarker signature.