靶向药物 z-bio 2024-12-09 678

　　DeepBlock is used for targeted drug design

　　With the support of National Natural Science Foundation of China projects (approval numbers: 62122025, U22A2037, 62202353), Professor Zeng Xiangxiang's team from Hunan University and Associate Professor Li Pengyong's team from Xi'an University of Electronic Science and Technology have made progress in using artificial intelligence to design targeted drug research. The related achievement, titled "A deep learning approach for rational ligand generation with toxicity control via reactive building blocks", was published on November 8, 2024 in Nature Computational Science. The paper link is: https://www.nature.com/articles/s43588-024-00718-0 .

　　In recent years, artificial intelligence technology, especially deep learning models, has attracted widespread attention in the field of drug design. However, due to the complexity of chemical space, existing methods are difficult to accurately capture reasonable drug molecular patterns, resulting in generated drugs often facing difficulties in synthesis and high toxicity. How to design rational drug molecules with controllable properties is still an urgent challenge that needs to be addressed.

　　Inspired by the DNA coding library (DEL) technology, the research team proposed a new deep learning method, DeepBlock (Figure), which targets protein sequences and generates ligands based on molecular blocks. The reactivity of molecular blocks is utilized to ensure the synthesizability of generated molecules, and a target aware molecular optimization method is proposed to control the toxicity of generated molecules. The experimental results showed that the synthesizability ratio of ligands generated by DeepBlock increased by 6% compared to the current best method, while maintaining high affinity and having higher drug like and molecular rationality, with an average drug like score of 0.54. In addition, while ensuring high affinity with the target protein, ligands with lower toxicity can be generated.

　　The molecular fragmentation and reconstruction algorithm, drug molecule generation algorithm, and drug optimization method proposed in this study not only provide innovative solutions to practical problems such as synthesis difficulty and toxicity control, but also provide new ideas for drug design, and provide strong guarantees for the safety and effectiveness of drugs.

。