Machine Learning Analysis of Molecular Indicators for Chronic Kidney Disease in Type 2 Diabetes

This study employs machine learning techniques to uncover insights from molecular research on diabetes, derived from laboratory experiments. We sourced our dataset from experiments conducted in a wet lab using Wistar Rats. These rats underwent specific treatments to induce a diabetic state. In our analysis, we quantified various trace elements within the molecular graph of the rat kidneys. Employing a specialized method, we gathered data on the proximal tubules, indicative of the absorption of water, sodium, and glucose into the bloodstream, and the distal tubules, which highlight the augmentation process to expel unnecessary fluids. Understanding these traces is crucial, as they can be indicative of chronic kidney diseases resulting from elevated blood sugar levels. To validate the robustness of our laboratory results, we utilized machine learning algorithms, specifically comparing the performance of Neural Networks, Stochastic Gradient Descent (SGD), and AdaBoost models. Additionally, we sought to understand the potential relationship between the positive/negative impacts of proximal and distal tubules. This relationship is key to discerning how Glucose Transporter 9 (GLUT9) and Sodium Glucose Cotransporter (SGLT2) might influence chronic kidney diseases. It's noteworthy that in the realm of type 2 diabetes, kidney diseases are often caused by the heightened expression of renal transporters like GLUT9 and SGLT2. Upon applying machine learning techniques, our results revealed that the AdaBoost model surpassed other models, achieving a precision value of 0.333 and an F1 value of 0.332.