Structure-based generative chemistry aims to explore much bigger chemical space to design a ligand with high binding affinity to the target proteins; it is a critical step inde novocomputer-aided drug discovery. Traditionalin silicomethods suffer from calculation inefficiency and the performances of existing machine learning methods could be bottlenecked by the auto-regressive sampling strategy. To address these concerns, we herein have developed a novel conditional deep generative model, PMDM, for 3D molecule generation fitting specified target proteins. PMDM incorporates a dual equivariant diffusion model framework to leverage the local and global molecular dynamics to generate 3D molecules in a one-shot fashion. By considering the conditioned protein semantic information and spatial information, PMDM is able to generate chemically and conformationally valid molecules which suitably fit pocket holes. We have conducted comprehensive experiments to demonstrate that PMDM can generate drug-like, synthesis-accessible, novel, and high-binding affinity molecules targeting specific proteins, outperforming the state-of-the-art (SOTA) models in terms of multiple evaluation metrics. In addition, we perform chemical space analysis for generated molecules and lead compound optimization for SARS-CoV-2 main protease (Mpro) by only utilizing three atoms as the seed fragment. The experimental results implicate that the structures of generated molecules are rational compared to the reference molecules, and PMDM can generate massive bioactive molecules highly binding to the targeted proteins which are not included in the training set.

Retrosynthetic planning plays an important role in the field of organic chemistry, which could generate a synthetic route for the target product. The synthetic route is a series of reactions which are started from the available molecules. The most challenging problem in the generation of the synthetic route is the large search space of the candidate reactions. Estimating the cost of candidate reactions has been proved effectively to prune the search space, which could achieve a higher accuracy with the same search iteration. And the estimation of one reaction is comprised of the estimations of all its reactants. So, how to estimate the cost of these reactants will directly influence the quality of results. To get a better performance, we propose a new framework, named GNN-Retro, for retrosynthetic planning problem by combining graph neural networks(GNN) and the latest search algorithm. The structure of GNN in our framework could incorporate the information of neighboring molecules, which will improve the estimation accuracy of our framework. The experiments on the USPTO dataset show that our framework could outperform the state-of-the-art methods with a large margin under the same settings.