Drug Target Interaction through Computational Approach

The development of effective drugs for diseases requires a comprehensive understanding of the binding affinity between diabetic ligands and target proteins at the molecular level. This study addresses the critical challenge of evaluating these interactions through the application of deep learning (DL) techniques, with a specific focus on predicting binding affinity. Leveraging molecular docking techniques, we employed deep learning DL models to streamline the analysis, providing a more efficient and accurate means of identifying potential drug candidates. Our research pioneers the use of deep learning in discerning molecular interactions and binding sites within proteins. By unraveling the complexities of protein-ligand interactions, our work contributes to the optimization of drug discovery efforts for diabetes. Drug Target Interaction aims to create safe and effective medicines, but it’s expensive and slow. During the "hit-to-lead" phase, promising compounds are found, but drugs interacting with multiple targets create challenges, causing side effects and lower effectiveness. We need new methods to make this phase easier and reduce risks from drugs affecting many targets and making the process faster. To validate the reliability of our predictions, we compared the predicted binding affinity with experimental data, demonstrating the technical proficiency of our approach. The integration of molecular docking techniques and the validation of predictions against experimental data contribute to the advancement of techniques for optimizing drug development in the context of diabetes. The significance of integrating Deep Learning (DL) into drug discovery. Our method efficiently predicts top drug-protein pairs, with predicted values closely aligning with actual values. Additionally, our approach facilitates drug repurposing by identifying drugs with high binding affinity for target proteins associated with diabetes. We calculated the AUROC metric values during the model training, achieving 0.85 and 0.82 for the training and validation sets, respectively. In the testing set, the ROC curve and the determined optimal threshold were utilized, resulting in an AUROC of 0.813.