Reproduce “Multi-objective fluorescent molecule design with a data-physics dual-driven generative framework” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yanheng Li, Zhichen Pu, Lijiang Yang, Zehao Zhou, Yi Qin Gao
The LUMOS framework introduces a data-physics dual-driven approach for the inverse design of fluorescent small molecules, integrating AI generative models with quantum chemical calculations. This system enables multi-objective optimization and de novo discovery of molecules with tailored optical and physicochemical properties, demonstrating superior performance in property prediction and generating novel cell-permeable fluorophores.
Key insights
- A continuous and semantic latent chemical space can effectively encode molecular structures, allowing for efficient navigation and preservation of chemical relationships.
- The integration of data-driven neural networks with GPU-accelerated quantum chemical calculations (TD-DFT) provides a robust strategy to balance speed and accuracy, addressing the limitations of each approach individually.
- Atomic attention weights in graph predictors can provide physically interpretable insights into structure-property relationships by aligning with frontier molecular orbital distributions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.