Identification of key biomolecules in adrenocortical cancer progression via bioinformatics and machine learning approaches

Adrenal Adenomas (ACA) are benign masses that form in the adrenal gland, while adrenocortical carcinoma (ACC) is a malignant aggressive tumor. In the present study, our aim is to investigate critical biomolecules from the onset of adrenocortical adenoma to the progression of carcinoma with bioinformatics and machine learning approaches. We will use a new analysis method to identify a statistically significant Spearman's gene correlation coefficient that have both a positive correlation in the case state and a negative correlation in the control state (or vice versa). Three RGEC networks were recontructed for ACA-Normal tissue (ACA-N), ACC-Normal tissue (ACC-N), and ACC-ACA and hub genes of RGEC networks were identified. PLA2G4A gene in ACA-N group, 14 genes (i.e: SAE1, ATRX, PNMT, RYBP, NCAPG2, PDK2, C7, RNF114, PTPRB, NPY1R, PARM1, PGK1, HMMR and ETC2) in ACC-N group, the FMO2 and UBE2S genes in the ACC-ACA group were determined. Diagnostic and prognostic performance of the resultant genes were performed through machine learning (ML) classification algorithms including K Neighbor Classifier, Logistic Regression, MLP Classifier, Decision Tree Classifier, Random Forest Classifier, Gradient Enhancement Classifier, Cat Enhancement Classifier, LGBM Classifier, and XGB Classifier. It was found that 14 central genes of RCEG network for ACC-N group efficiently discriminate the ACC tissues from the normal tissues compared to the performance of discriminating in the live and dead specimens. The discovery of these putative biomolecules may also be an important step in preventing a benign tumor from turning into a malignant one.