Drug Molecule Generation Method Based on Fusion of Protein Sequence Features

Computer-aided drug design (CADD) plays a crucial role in accelerating the drug discovery process, particularly when structural information of the target is unavailable. This paper proposed a drug design model based on the protein sequence features. The model utilizes Transformer and a graph neural network (GNN) algorithm based on the weighted protein graph to obtain protein sequence features. This approach allows model to effectively understand and utilize complex information about protein sequences and their spatial structure. By fusing these features through an attentional mechanism, it is used as a constraint to guide the generation of drug molecules. In order to comprehensively evaluate and screen out the most promising molecular candidates, this paper applies the non-dominated sorting algorithm (NSGA) to sort various drug properties of newly generated drug molecules. Finally, the newly generated drug molecules were evaluated for quality and molecular docking. The results showed that the proposed method is capable of generating drug molecules with high affinity for the target and optimized binding conformation.