Integrative approach: Advancing endometrial cancer biomarkers through machine learning and network-derived RNA-Seq screening

Abstract

The study aims to identify potential RNA biomarkers for progressive endometrial cancer with a machine learning-based approach. High throughput screening of RNA-Seq expression data identified Differentially Expressed Genes(DEGs) for various stages (Stage I-Stage III) of endometrial cancer. Overlapping genes among different stages of the disease were screened and compared for commonality with TCGA and the GTEx RNA Sequencing expression data. 1466 genes were found overlapping and expressed differentially in the progressive disease from healthy endometrial to Stage I to Stage II to Stage III endometrial cancer. Machine learning (ML) classifiers, namely Random Forest, Principal Component Analysis, Multinomial Naïve Bayes, and Support Vector Machine with Recursive Feature Elimination, helped select the most important ranked interacting DEGs. Overlapping important ranked DEGs, hub proteins, and significant modules from protein-protein interaction network analysis were used as candidate biomarkers for progressive endometrial cancer. In silico validations of druggable top-ranked cancer targets for their essentiality on cancer cell lines and high expression in immunohistochemistry analysis were performed. Further, their dysregulated expressions were evaluated in different cancers. The dysregulation of the proposed biomarkers, namely PKM, RAN, PHGDH, and SLC7A5, showed a decrease in the survival of the patients associated with endometrial cancer. These overlapping associations offer common cancer-specific targets among different stages of cancer which may assist, during the onset of the disease, in pointing to proliferating cancer cells. They provide significant visions in the understanding of progressive endometrial cancer.