This is a page that curates AI-related papers published worldwide. All content here is summarized using Google Gemini and operated on a non-profit basis. Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.
Controllable 3D Molecular Generation for Structure-Based Drug Design Through Bayesian Flow Networks and Gradient Integration
Created by
Haebom
Author
Seungyeon Choi, Hwanhee Kim, Chihyun Park, Dahyeon Lee, Seungyong Lee, Yoonju Kim, Hyoungjoon Park, Sein Kwon, Youngwan Jo, Sanghyun Park
Outline
This paper identifies the limitations of generative models used in existing structure-based drug design (SBDD) for generating 3D molecules and proposes CByG, a novel framework to overcome these limitations. While existing diffusion-based generative models primarily focus on target protein binding affinity for evaluation, synthesizability and selectivity are also important factors in real-world drug discovery. CByG is a gradient-based conditional generative model that extends Bayesian flow networks. It effectively integrates various pharmacological characteristics, such as binding affinity, synthesizability, and selectivity, to guide molecule generation. Experimental results demonstrate that CByG outperforms existing models across multiple evaluation criteria, suggesting its applicability to real-world drug discovery.
Takeaways, Limitations
•
Takeaways:
◦
A new generative model framework (CByG) is presented to overcome the limitations of existing SBDD.
◦
Proposing a comprehensive evaluation method that considers not only binding affinity but also synthetic feasibility and selectivity.
◦
Effectively integrating various pharmacological properties suggests potential contribution to actual drug discovery.
◦
Experimentally proven that CByG outperforms existing models.
•
Limitations:
◦
Further research is needed on the proposed assessment method and the generalizability of CByG.
◦
Further validation and optimization are needed for application to actual drug development processes.
◦
Analysis of the computational cost and complexity of the CByG model is needed.
