DeepMind AI Solves 50-Year-Old Protein Folding Challenge
DeepMind's AlphaFold 2 software accurately predicts 3D protein structures from DNA sequences, potentially revolutionizing drug discovery. The breakthrough solves a major 50-year-old challenge in molecular biology.
Researchers at DeepMind achieve a major scientific milestone by creating an AI system that accurately predicts the three-dimensional structure of proteins from DNA sequences. The software, known as AlphaFold 2, solves a 50-year-old challenge in molecular biology by determining these complex shapes to within an atom's width of accuracy.
The DeepMind team demonstrates this breakthrough at a biennial algorithm competition called CASP, where AlphaFold 2 vastly outperforms all other existing methods. Across more than 100 tested proteins, the AI achieves atomic-level accuracy in two-thirds of the cases and shows high accuracy in almost all of the remaining tests.
Scientists praise the computational advance as a stunning development that holds massive potential for revolutionizing the hunt for new medicines. DeepMind currently plans to figure out how to provide scalable access to this powerful system for academic researchers and potential commercial partners in the pharmaceutical industry.