.svg){kind=spacer}
Details
- Microsoft's AI for Good Lab has released Seq2Symm, an AI tool capable of predicting protein assembly symmetry at a rate of 80,000 proteins per hour through the ESM2 foundation model.
- The model was developed with leading researchers, including Nobel Prize winner David Baker, MIT’s Bonnie Berger, and Gregory Bowman from the University of Pennsylvania.
- Seq2Symm shows improved performance over traditional template-based methods, reaching AUC-PR scores of 0.44–0.49 versus 0.23–0.25 for older techniques.
- It is designed for integration with existing protein structure prediction tools, such as AlphaFold2-multimer and RoseTTAFold2, enabling more comprehensive 3D modeling efforts.
- The tool is open-source and has been applied to five entire proteomes, spanning 3.5 million previously unlabeled protein sequences.
Impact
Seq2Symm streamlines protein research, enabling faster breakthroughs in areas like drug discovery, disease research, and synthetic biology. By making this technology open-source, Microsoft empowers broader scientific collaboration and accelerates progress in the life sciences. The release reinforces Microsoft's commitment to open AI innovation in biotechnology, filling key gaps in existing computational approaches.
