INTEGRATING RNA-SEQ ANALYSIS WITH MACHINE LEARNING APPROACHES FOR THE IDENTIFICATION OF HUB GENES AFFECTED BY EXPOSURE OF BISPHENOL-A IN PATIENTS SUFFERING FROM TYPE 2 DIABETES

Abstract

Type 2 Diabetes (T2D) is a metabolic disorder caused (Butler, Janson and Bonner-Weir) due to development of insensitivity of insulin (Butler, Janson and Bonner-Weir) and loss of β-cell functioning (Butler, Janson and Bonner-Weir) (Iqra Hameed) (Vito.) (Rorsmam.). Evidence suggests that exposure to bisphenol-A (BPA) may contribute in the development and progression of type 2 diabetes. Thus, this study integrates bioinformatics and machine learning approaches to identify key genes and pathways involved linking BPA with T2DM. Datasets related to T2D were retrieved from GEO, which were then pre-processed, normalized and batch corrected. DEG analysis was then done to identify the up and down regulating genes. Following of which the data was split into training and test set. BPA related genes were then retrieved from database comparative toxicogenomics database, swiss target prediction and STITCH. While, disease related genes were retrieved from GeneCards, MalaCards and OMIM. Results from these were then overlapped to identify hub genes which were assessed using machine learning approaches RF, SVM and LASSO to identify potential candidate biomarkers. Next, ROC curve analysis was done on the genes i.e ABCG1, AMPD2 and HBA2. This integrated approach also did pathway enrichment of the potential diagnostic biomarkers which highlighted pathways related to metabolic regulation, mitochondrial function and inflammatory responses. Lastly, molecular docking was used for assessing the binding interaction occurring between BPA and ABCG1, AMPD2 and HBA2 which yielded a binding affinity of -7.4, -8.1 and – 8.1 kcal/mol respectively. Overall this study helped in identifying a potential link between BPA and T2DM and in identifying the potential candidates that may contribute progression of disease.