Protein-Folding AI Breakthrough: ESMFold2 Expands the Known Protein Universe
### A Quantum Leap in Protein Mapping: ESMFold2 Adds 150 Million Structures The European Molecular Biology Laboratory (EMBL) has unveiled an open‑source atlas generated by its new protein‑folding engine, ESMFold2. Leveraging transformer models trained on billions of amino‑acid sequences, the system can predict three‑dimensional structures in seconds, expanding the publicly available catalogue by more than 150 million entries—far surpassing the scale of AlphaFold’s original database. #### Key Takeaways - **Massive expansion:** Over 150 million newly predicted protein structures are now publicly accessible. - **Speed and scalability:** ESMFold2 delivers high‑confidence predictions within seconds, thanks to transformer‑based architecture. - **Open‑source impact:** The atlas is freely available, encouraging widespread adoption across academia and industry. - **Scientific breadth:** The model was trained on billions of sequences, enhancing coverage of previously uncharacterized protein families. - **Strategic advantage:** By dwarfing AlphaFold’s initial repository, ESMFold2 positions EMBL at the forefront of structural biology and AI‑driven discovery. #ProteinFolding #ESMFold2 #StructuralBiology #AIResearch #EMBL #Bioinformatics #ProteinStructure #DeepLearning #Genomics #newsababil360 [Read Full Article](https://news.ababil360.com/protein-folding-ai-breakthrough-esmfold2-expands-the-known-protein-universe/)
