A novel deep learning system, detailed in the _International Journal of Reasoning-based Intelligent Systems_, promises to accelerate drug discovery by overcoming the vast number of potential drug-like molecules that cannot be practically tested. This system treats drug candidates as graphs, with atoms as nodes and bonds as edges, and simultaneously processes their SMILES strings, a text-based chemical structure representation. By combining these structural and sequential data, the model significantly improves performance, achieving a score of 0.889 on public benchmarks, indicating its effectiveness in distinguishing active from inactive compounds. This advancement offers a substantial speed increase, screening one million molecules in just fifteen minutes, which is 80% faster than traditional virtual screening methods, potentially unblocking industry bottlenecks in the costly and time-consuming process of bringing new pharmaceuticals to market.