The paper introduces DECOMPDiff, a novel diffusion model designed for structure-based drug design (SBDD) that decomposes ligand molecules into arms and scaffold regions. This decomposition is inspired by conventional pharmaceutical practices, where different parts of the ligand have distinct roles in binding to the target. The model incorporates decomposed priors over arms and scaffold to guide the generation process, improving the efficiency and effectiveness of molecular generation. Additionally, the model includes bond diffusion to simultaneously generate atoms and bonds, enhancing the realism and stability of the generated molecules. Extensive experiments on the CrossDocked2020 dataset demonstrate that DECOMPDiff achieves state-of-the-art performance in generating high-affinity molecules, with an Avg. Vina Dock score of −8.39 and a Success Rate of 24.5%. The method's effectiveness is further validated through ablation studies, which show that each component of the model contributes significantly to the overall performance.The paper introduces DECOMPDiff, a novel diffusion model designed for structure-based drug design (SBDD) that decomposes ligand molecules into arms and scaffold regions. This decomposition is inspired by conventional pharmaceutical practices, where different parts of the ligand have distinct roles in binding to the target. The model incorporates decomposed priors over arms and scaffold to guide the generation process, improving the efficiency and effectiveness of molecular generation. Additionally, the model includes bond diffusion to simultaneously generate atoms and bonds, enhancing the realism and stability of the generated molecules. Extensive experiments on the CrossDocked2020 dataset demonstrate that DECOMPDiff achieves state-of-the-art performance in generating high-affinity molecules, with an Avg. Vina Dock score of −8.39 and a Success Rate of 24.5%. The method's effectiveness is further validated through ablation studies, which show that each component of the model contributes significantly to the overall performance.