AlphaFold 3 can predict the structures and interactions of nearly all of life's molecules with state-of-the-art accuracy including proteins, DNA and RNA. It's hoped that the system can be used to design new drugs and treatments faster and more accurately than current methods.
AI model provides a 50 percent improvement over existing prediction methods on PoseBusters benchmark, doubling prediction accuracy in certain categories of molecular interactions. AlphaFold 3 extends predictions beyond proteins to DNA, RNA, and small molecule ligands.
AlphaFold 3 has a 4.4 percent error rate in predicting 'chirality' and occasionally produces overlapping atoms. Despite these shortcomings, the AI model represents significant progress in predicting molecular structures.
[
add
]
[
|
|
...
]