Compared to the previous drug discovery methods which develop selective drug candidates with high affinities with a single specific target protein, the new drug discovery paradigm against multiple targets could treat complex disorders by modulating multiple targets to accomplish desired molecular responses. However, the previous works do not incorporate protein information and just learn the molecule data distributions, which are far away from the clinical application. Although there are previous works which focus on dual targets, however, they require a large amount of molecule data that are bioactive toward the specific dual targets to train the model and do not incorporate protein information directly. These methods cannot be generalized to arbitrary dual targets. In this article, we have constructed the dual-target bioactive molecule dataset for model training and evaluation. More importantly, we proposed a novel generative model based on a discrete diffusion framework (DiffDTM) to...

Advances in deep generative models shed light on de novo molecule generation with desired properties. However, molecule generation targeted for dual protein targets still faces formidable challenges including insufficient protein 3-D structure data requisition for conditioned model training, inflexibility of auto-regressive sampling, and model generalization to unseen targets. Here, this study proposed diffusion model for dual targets-based molecule generation (DiffDTM), a novel unified structure-free deep generative framework based on a diffusion model for dual-target based molecule generation to address the above issues. Specifically, DiffDTM receives representations of protein sequences and molecular graphs pretrained on large-scale datasets as inputs instead of protein and molecular conformations and incorporates an information fusion module to achieve conditional generation in a one-shot manner. We perform comprehensive multiview experiments to demonstrate that DiffDTM can generat...