MIT Open Source Model CheMeleon Predicts Molecular Behavior with Excellence
The story so far
MIT's open-source model CheMeleon, pre-trained with millions of molecular data, can predict 1600 chemical descriptor properties of new molecules, outperforming other baseline and expert models in 58 test cases. The model has been adopted by several pharmaceutical companies and has performed outstandingly in the OpenADMET competition, significantly reducing the need for experimental validation of new molecules.
AI-written from the reports · Updated 12 days ago
Coverage timeline
Follow the coverage to see every side of the story.
- MIT News · AIHow to predict the behavior of molecules from scant data
MIT's open-source model CheMeleon, pre-trained with millions of molecular data, can predict 1600 chemical descriptor properties of new molecules, outperforming other baseline and expert models in 58 test cases. The model has been adopted by several pharmaceutical companies and has performed outstandingly in the OpenADMET competition, significantly reducing the need for experimental validation of new molecules.