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.
Compositional Flows for 3D Molecule and Synthesis Pathway Co-design
Created by
Haebom
Author
Tony Shen, Seonghwan Seo, Ross Irwin, Kieran Didi, Simon Olsson, Woo Youn Kim, Martin Ester
Outline
Compositional Generative Flows (CGFlow) is a novel framework that extends flow matching to generate component objects with continuous features. CGFlow formalizes compositional state transitions as a simple extension of the flow matching interpolation process, enabling reward-based compositional sampling by leveraging the theoretical foundation of GFlowNets. It is applied to synthesizable drug design by jointly designing the synthetic trajectories and 3D binding poses of molecules, achieving state-of-the-art binding affinities on all 15 targets in the LIT-PCBA benchmark and improving sampling efficiency by 5.8x over 2D synthesis-based baselines. It is also the first method to achieve state-of-the-art performance on both Vina Dock (-9.38) and AiZynth success rate (62.2%) on the CrossDocked benchmark.
Takeaways, Limitations
•
Takeaways:
◦
An efficient framework for generating constituent objects with continuous features (CGFlow).
◦
Generation of efficient configuration structures via reward-based sampling based on GFlowNets.
◦
Achieving state-of-the-art performance in synthetic drug design (LIT-PCBA and CrossDocked benchmarks).
◦
5.8x improvement in sampling efficiency compared to 2D synthesis-based methods.
◦
Achieving cutting-edge performance in Vina Dock and AiZynth success rates.
•
Limitations:
◦
The paper does not specifically mention Limitations. Additional experiments or analyses are needed to identify Limitations.
◦
Focused on a specific application area (drug design), further research is needed to determine generalizability to other fields.
