AlphaFold2
A protein structure prediction model developed by DeepMind, using attention mechanisms as its core computational component. AlphaFold2 achieved landmark accuracy in predicting three-dimensional protein structures from amino acid sequences, representing a significant advance in computational biology.
Core Facts
AlphaFold是DeepMind于2020年推出的蛋白质结构预测系统
AlphaFold2 overwhelmingly surpassed all traditional methods at the CASP14 competition, with prediction accuracy approaching the level of experimental measurement.
AlphaFold is an AI model applied to protein structure prediction, representing the application of AI in scientific research at the level of specific specialized models.
Timeline (last 90 days)
AlphaFold2 achieved a median score of about 90 GDT-TS in the 2020 CASP14 assessment, far exceeding the previous level of about 60 points.
AlphaFold2 surpassed traditional methods by an overwhelming margin at the CASP competition in 2020.
AlphaFold2 was released by DeepMind in 2021 and can predict protein three-dimensional structures with near-atomic accuracy.
AlphaFold2 is essentially a bimodal sequence-structure model
Microsoft has the "AI for Health" program, Google has the "DeepMind Science Team," and Meta developed the ESM protein language model.
AlphaFold2's breakthrough advances in protein structure prediction have significantly lowered the barrier to structural biology, enabling researchers with a medicinal chemistry background to more quickly conduct structure-based drug design (SBDD).
AlphaFold2 compressed the process of resolving protein structures—which could take years with X-ray crystallography or cryo-electron microscopy—down to a matter of minutes.
AlphaFold2 won CASP14 with a median GDT score of 92.4, approaching the level of experimental methods such as X-ray crystallography and cryo-electron microscopy.
DeepMind's Demis Hassabis publicly stated after AlphaGo defeated Lee Sedol in 2016 that AI would achieve breakthroughs on scientific problems such as protein folding, a judgment that was validated four years later by AlphaFold2.
AlphaFold2 solved the protein structure prediction problem in 2020, achieving experimental-level prediction accuracy at the CASP14 competition with a median GDT score exceeding 92.
3 more timeline events
All Facts (18)
AlphaFold是DeepMind于2020年推出的蛋白质结构预测系统
90%VerifiedAlphaFold2 overwhelmingly surpassed all traditional methods at the CASP14 competition, with prediction accuracy approaching the level of experimental measurement.
90%VerifiedAlphaFold is an AI model applied to protein structure prediction, representing the application of AI in scientific research at the level of specific specialized models.
80%VerifiedAlphaFold2 compressed the process of resolving protein structures—which could take years with X-ray crystallography or cryo-electron microscopy—down to a matter of minutes.
75%VerifiedAlphaFold solved the protein folding problem in 2020.
75%VerifiedAlphaFold solved the protein folding problem that had puzzled the biology community for 50 years—namely, predicting the three-dimensional structure of a protein from its amino acid sequence—through deep learning technology.
70%UnverifiedDeepMind公开了超过2亿个蛋白质的预测结构数据库，覆盖了几乎所有已知蛋白质
70%UnverifiedAlphaFold2 achieved a median score of about 90 GDT-TS in the 2020 CASP14 assessment, far exceeding the previous level of about 60 points.
50%UnverifiedAlphaFold2 surpassed traditional methods by an overwhelming margin at the CASP competition in 2020.
50%UnverifiedAlphaFold2 was released by DeepMind in 2021 and can predict protein three-dimensional structures with near-atomic accuracy.
50%UnverifiedAlphaFold2 is essentially a bimodal sequence-structure model
50%UnverifiedMicrosoft has the "AI for Health" program, Google has the "DeepMind Science Team," and Meta developed the ESM protein language model.
50%UnverifiedAlphaFold2's breakthrough advances in protein structure prediction have significantly lowered the barrier to structural biology, enabling researchers with a medicinal chemistry background to more quickly conduct structure-based drug design (SBDD).
50%UnverifiedAlphaFold2 won CASP14 with a median GDT score of 92.4, approaching the level of experimental methods such as X-ray crystallography and cryo-electron microscopy.
50%UnverifiedDeepMind's Demis Hassabis publicly stated after AlphaGo defeated Lee Sedol in 2016 that AI would achieve breakthroughs on scientific problems such as protein folding, a judgment that was validated four years later by AlphaFold2.
50%UnverifiedAlphaFold2 solved the protein structure prediction problem in 2020, achieving experimental-level prediction accuracy at the CASP14 competition with a median GDT score exceeding 92.
50%UnverifiedAlphaFold2 emerged in 2021, dramatically improving protein structure prediction accuracy; RFdiffusion was developed by David Baker's team and uses a diffusion model to directly sample in structure space.
50%UnverifiedAlphaFold2, based on multiple sequence alignment (MSA) and the Evoformer attention architecture, can directly predict three-dimensional atomic coordinates from protein amino acid sequences, and stunned the scientific community in 2020 with performance approaching experimental accuracy.
50%