AlphaFold 2 solves protein structure prediction
- machine learning
DeepMind’s AlphaFold 2 achieved a major improvement in protein-structure prediction at the CASP14 assessment in 2020. The system used learned representations and structural reasoning to predict three-dimensional protein arrangements from sequence-related information with high accuracy on many benchmark targets. The result transformed a longstanding scientific-computing problem, but “solves” needs qualification: proteins can have multiple states, interactions and dynamic behavior, and confidence varies. A predicted structure is not automatically a complete experimental account of biological function. The benchmark milestone should also be distinguished from later software and database releases. *Image: Wikimedia Commons (Demis Hassabis in Stockholm, 2024; photo by John Sears; CC BY-SA 4.0).* **Timeline anchor:** 30 November 2020 CE. The date CASP14's results were announced. **References:** [Source 1](https://en.wikipedia.org/wiki/AlphaFold).
- Date precision
- The date CASP14's results were announced.
Sources
Related
On other timelines
- AlphaFold reports accelerator-based protein-structure prediction on GPU and ASIC programming · parallel hardware, standards and applications by @athan
- AlphaFold2 demonstrates highly accurate predictions for many protein structures on Studying life · history of biology, evidence and discovery by @athan
Open AlphaFold 2 solves protein structure prediction in the timeline