The application of Graph Neural Networks (GNNs) in computational chemistry provides a powerful approach to modeling and predicting the properties of molecular compounds. GNNs represent atoms as nodes and bonds as edges, capturing the complex interactions within molecular graphs. This approach offers a robust method for predicting chemical properties, including molecular stability, reactivity, and toxicity. In this paper, we explore various GNN architectures and their ability to generalize across different molecular datasets, such as QM9 and MoleculeNet. As a specific application, we propose a novel framework that utilizes GNNs to predict and identify potential HIV inhibitor molecules by analyzing their graph-based representations. This research aims to contribute to the discovery and design of effective HIV inhibitors, offering a promising direction for future antiviral drug development.