Conception et identification de molécules inhibitrices des kinases : voie de développement de médicaments anti-cancéreux

Abstract

Among recent cancer treatment strategies, targeted therapies play a central role. These aim to halt tumor progression by inhibiting dysregulated protein kinases whose overexpression is directly linked to cancer cell proliferation. In this thesis work, we employed several computational methods to improve predictive accuracy in modeling Quantitative Structure–Activity Relationships (QSARs) related to kinase inhibition. A new approach based on a hybrid eXtreme Gradient Boosting (XGBoost) and Deep Neural Network (DNN) architecture was introduced to enhance QSAR prediction accuracy. Furthermore, we used Ligand-Based Virtual Screening IV(LBVS) and rigorous Structure-Based Drug Design (SBDD) tools to identify novel, potentially kinase-inhibiting molecules. These compounds show promising moderate to high potency based on molecular docking and molecular dynamics simulation analyses. Our results suggest, on one hand, a significant advance in theoretical modeling, and on the other hand, reveal the anticancer potential of certain molecules.