AlphaFold: Difference between revisions

From Proteopedia
Jump to navigationJump to search
← Older editNewer edit →
Joel L. Sussman (talk | contribs)
ConfirmAccount, Editor, Mutation, Print3D, Tester, Bureaucrats, journal, tester
11,710 edits
No edit summary
Joel L. Sussman (talk | contribs)
ConfirmAccount, Editor, Mutation, Print3D, Tester, Bureaucrats, journal, tester
11,710 edits
No edit summary
Line 1: Line 1:
In 2020, the '''AlphaFold2'''<ref name="senior202001">PMID: 31942072</ref><ref name="alphafoldwikipedia">[https://en.wikipedia.org/wiki/AlphaFold AlphaFold] at Wikipedia.</ref> system of [http://deepmind.com DeepMind]<ref name="deepmindblog">[https://deepmind.com/blog/article/alphafold-a-solution-to-a-50-year-old-grand-challenge-in-biology AlphaFold: a solution to a 50-year-old grand challenge in biology], DeepMind Blog, November 30, 2020.</ref><ref name="deepmindwikipedia">[https://en.wikipedia.org/wiki/DeepMind DeepMind] at Wikipedia.</ref> demonstrated a '''major breakthrough'''<ref name="alquraishi">[https://moalquraishi.wordpress.com/2020/12/08/alphafold2-casp14-it-feels-like-ones-child-has-left-home/ AlphaFold2 @ CASP14: “It feels like one’s child has left home.”] by Mohammed AlQuraishi, December 8, 2020.</ref><ref name="casppressrelease">[https://predictioncenter.org/casp14/doc/CASP14_press_release.html Artificial intelligence solution to a 50-year-old science challenge could ‘revolutionise’ medical research], CASP Press Release, November 30, 2020.</ref><ref name="callaway" /><ref name="helliwell">[https://www.iucr.org/news/newsletter/volume-28/number-4/deepmind-and-casp14 DeepMind and CASP14] by John R. Helliwell, International Union of Crystallography Newsletter, December 4, 2020.</ref>. At [[Theoretical_models#2020:_CASP_14|CASP14]], AlphaFold2 was far better able, among over 100 competing groups, to '''predict structures, including sidechain positions''', so close to the subsequently revealed X-ray crystallographic structures as to differ by little more than the differences between two independently-determined X-ray structures of the same molecule. It did this for about two-thirds of the targets in the competition. AlphaFold2 has been hailed as '''largely solving the protein structure prediction problem for single-chain proteins'''<ref name="alquraishi" /><ref name="casppressrelease" /><ref name="callaway">PMID: 33257889</ref><ref name="helliwell" />. "Never in my life had I expected to see a scientific advance so rapid." said Mohammed AlQuraishi of Columbia University<ref name="alquraishi" />.
In 2020, the '''AlphaFold2'''<ref name="senior202001">PMID: 31942072</ref><ref name="alphafoldwikipedia">[http://en.wikipedia.org/wiki/AlphaFold AlphaFold] at Wikipedia.</ref> system of [http://deepmind.com DeepMind]<ref name="deepmindblog">[http://deepmind.com/blog/article/alphafold-a-solution-to-a-50-year-old-grand-challenge-in-biology AlphaFold: a solution to a 50-year-old grand challenge in biology], DeepMind Blog, November 30, 2020.</ref><ref name="deepmindwikipedia">[http://en.wikipedia.org/wiki/DeepMind DeepMind] at Wikipedia.</ref> demonstrated a '''major breakthrough'''<ref name="alquraishi">[http://moalquraishi.wordpress.com/2020/12/08/alphafold2-casp14-it-feels-like-ones-child-has-left-home/ AlphaFold2 @ CASP14: “It feels like one’s child has left home.”] by Mohammed AlQuraishi, December 8, 2020.</ref><ref name="casppressrelease">[http://predictioncenter.org/casp14/doc/CASP14_press_release.html Artificial intelligence solution to a 50-year-old science challenge could ‘revolutionise’ medical research], CASP Press Release, November 30, 2020.</ref><ref name="callaway" /><ref name="helliwell">[http://www.iucr.org/news/newsletter/volume-28/number-4/deepmind-and-casp14 DeepMind and CASP14] by John R. Helliwell, International Union of Crystallography Newsletter, December 4, 2020.</ref>. At [[Theoretical_models#2020:_CASP_14|CASP14]], AlphaFold2 was far better able, among over 100 competing groups, to '''predict structures, including sidechain positions''', so close to the subsequently revealed X-ray crystallographic structures as to differ by little more than the differences between two independently-determined X-ray structures of the same molecule. It did this for about two-thirds of the targets in the competition. AlphaFold2 has been hailed as '''largely solving the protein structure prediction problem for single-chain proteins'''<ref name="alquraishi" /><ref name="casppressrelease" /><ref name="callaway">PMID: 33257889</ref><ref name="helliwell" />. "Never in my life had I expected to see a scientific advance so rapid." said Mohammed AlQuraishi of Columbia University<ref name="alquraishi" />.


*See [[Theoretical_models#2020:_CASP_14]] for more about the initial demonstration at CASP14, and the reactions to it.
*See [[Theoretical_models#2020:_CASP_14]] for more about the initial demonstration at CASP14, and the reactions to it.
Line 7: Line 7:
==AlphaFold Database of Predictions==
==AlphaFold Database of Predictions==


In July, 2021, DeepMind made available over 300,000 structure predictions from amino acid sequences in their free [https://alphafold.ebi.ac.uk/ AlphaFold DB]<ref name="deepminddb">[https://deepmind.com/research/case-studies/alphafold#a_treasure_trove We’ve made AlphaFold predictions freely available to anyone in the scientific community] at DeepMind.com (date of release not specified, approximately July 2021).</ref><ref name="afdbebi">[https://www.ebi.ac.uk/pdbe/about/news/alphafold%E2%80%99s-protein-structure-predictions-now-available-explore AlphaFold’s protein structure predictions now available to explore] at the European Bioinformatics Institute, July 23, 2021.</ref><ref name="impacts">[https://www.embl.org/news/science/alphafold-potential-impacts/ Great expectations – the potential impacts of AlphaFold DB] at the European Bioinformatics Institute, July 22, 2021</ref><ref name="human">[https://www.embl.org/news/science/alphafold-database-launch/ DeepMind and EMBL release the most complete database of predicted 3D structures of human proteins] at the European Bioinformatics Institute, July 22, 2021.</ref>. These predictions include nearly all ~20,000 proteins in the human proteome, 36% with very high confidence, and another 22% with high confidence<ref name="human" /><ref name="human-nature">PMID: 34293799</ref>. Also included are ''E. coli'', fruit fly, mouse, zebrafish, malaria parasite and tuberculosis bacteria<ref name="human" />. Limitations of these predictions were enumerated<ref name="impacts" />, including:
In July, 2021, DeepMind made available over 300,000 structure predictions from amino acid sequences in their free [http://alphafold.ebi.ac.uk/ AlphaFold DB]<ref name="deepminddb">[http://deepmind.com/research/case-studies/alphafold#a_treasure_trove We’ve made AlphaFold predictions freely available to anyone in the scientific community] at DeepMind.com (date of release not specified, approximately July 2021).</ref><ref name="afdbebi">[http://www.ebi.ac.uk/pdbe/about/news/alphafold%E2%80%99s-protein-structure-predictions-now-available-explore AlphaFold’s protein structure predictions now available to explore] at the European Bioinformatics Institute, July 23, 2021.</ref><ref name="impacts">[http://www.embl.org/news/science/alphafold-potential-impacts/ Great expectations – the potential impacts of AlphaFold DB] at the European Bioinformatics Institute, July 22, 2021</ref><ref name="human">[http://www.embl.org/news/science/alphafold-database-launch/ DeepMind and EMBL release the most complete database of predicted 3D structures of human proteins] at the European Bioinformatics Institute, July 22, 2021.</ref>. These predictions include nearly all ~20,000 proteins in the human proteome, 36% with very high confidence, and another 22% with high confidence<ref name="human" /><ref name="human-nature">PMID: 34293799</ref>. Also included are ''E. coli'', fruit fly, mouse, zebrafish, malaria parasite and tuberculosis bacteria<ref name="human" />. Limitations of these predictions were enumerated<ref name="impacts" />, including:
* Inability to predict protein-protein or protein-DNA/RNA/ligand complexes. [[#RoseTTAFold]] claims to have made progress on this.
* Inability to predict protein-protein or protein-DNA/RNA/ligand complexes. [[#RoseTTAFold]] claims to have made progress on this.
* Does not predict ligands, cofactors, metals, ions, glycosylation, etc.
* Does not predict ligands, cofactors, metals, ions, glycosylation, etc.
Line 34: Line 34:
===RoseTTAFold===
===RoseTTAFold===


Also in July, 2021, Minkyung Baek and a large team in the [https://www.bakerlab.org/ group of David Baker] published their  '''RoseTTAFold''' employing a three-track nework, based in part on methods inspired by AlphaFold but not yet fully-detailed by DeepMind. They reported "accuracies approaching those of DeepMind in CASP14"<ref name="baek1">PMID: 34282049</ref>. At the time of its release in July, 2021, it had outperformed all other available structure prediction ''servers''<ref name="baek1" />.
Also in July, 2021, Minkyung Baek and a large team in the [http://www.bakerlab.org/ group of David Baker] published their  '''RoseTTAFold''' employing a three-track nework, based in part on methods inspired by AlphaFold but not yet fully-detailed by DeepMind. They reported "accuracies approaching those of DeepMind in CASP14"<ref name="baek1">PMID: 34282049</ref>. At the time of its release in July, 2021, it had outperformed all other available structure prediction ''servers''<ref name="baek1" />.


The [https://robetta.bakerlab.org/ RoseTTAFold Server] was made freely available. (Open the ''Structure Prediction'' menu at the top and choose ''Submit''. At the form, be sure to check ''RoseTTAFold'' before submitting your job).
The [http://robetta.bakerlab.org/ RoseTTAFold Server] was made freely available. (Open the ''Structure Prediction'' menu at the top and choose ''Submit''. At the form, be sure to check ''RoseTTAFold'' before submitting your job).


===AlphaFold Colab===
===AlphaFold Colab===
Google provides "Colaboratories" (Colabs). A Colab "allows anybody to write and execute arbitrary python code through the browser, and is especially well suited to machine learning, data analysis and education"<ref name="colabfaq">[https://research.google.com/colaboratory/faq.html Collaboratory FAQ] at Google.</ref>.
Google provides "Colaboratories" (Colabs). A Colab "allows anybody to write and execute arbitrary python code through the browser, and is especially well suited to machine learning, data analysis and education"<ref name="colabfaq">[http://research.google.com/colaboratory/faq.html Collaboratory FAQ] at Google.</ref>.


DeepMind has provided an [https://colab.research.google.com/github/deepmind/alphafold/blob/main/notebooks/AlphaFold.ipynb Alphafold Colab] that uses a "slightly simplified" version of AlphaFold version 2.0: "While accuracy will be near-identical to the full AlphaFold system on many targets, a small fraction have a large drop in accuracy due to the smaller MSA and lack of templates.". The AlphaFold Colab is '''free to use'''. The code is executed in a virtual machine private to your account, and data are stored on Google Drive. Nothing is installed on your computer; "everything happens in the cloud on Google Colab"<ref name="alphafoldcolab">[https://colab.research.google.com/github/deepmind/alphafold/blob/main/notebooks/AlphaFold.ipynb Alphafold Colab].</ref>
DeepMind has provided an [http://colab.research.google.com/github/deepmind/alphafold/blob/main/notebooks/AlphaFold.ipynb Alphafold Colab] that uses a "slightly simplified" version of AlphaFold version 2.0: "While accuracy will be near-identical to the full AlphaFold system on many targets, a small fraction have a large drop in accuracy due to the smaller MSA and lack of templates.". The AlphaFold Colab is '''free to use'''. The code is executed in a virtual machine private to your account, and data are stored on Google Drive. Nothing is installed on your computer; "everything happens in the cloud on Google Colab"<ref name="alphafoldcolab">[http://colab.research.google.com/github/deepmind/alphafold/blob/main/notebooks/AlphaFold.ipynb Alphafold Colab].</ref>


For those unfamiliar with Colabs, the user interface may look unfamilar, but the instructions are clear and straightforward to use. The mentions of "Runtime -> Run after" refer to the Runtime pull-down menu at the very top of the page. Getting a result may take several hours.
For those unfamiliar with Colabs, the user interface may look unfamilar, but the instructions are clear and straightforward to use. The mentions of "Runtime -> Run after" refer to the Runtime pull-down menu at the very top of the page. Getting a result may take several hours.
Line 51: Line 51:


* '''AlphaFold protein structure predictions - a step change for biology.'''
* '''AlphaFold protein structure predictions - a step change for biology.'''
: (Report by Oana Stroe, Senior Communications Officer at EMBL-EBI. 28 July 2021 at [https://bit.ly/2UZREPx FEBS Network])
: (Report by Oana Stroe, Senior Communications Officer at EMBL-EBI. 28 July 2021 at [http://bit.ly/2UZREPx FEBS Network])
: Sameer Velankar and Gerard Kleywegt, from the Protein Data Bank in Europe, and Alex Bateman, Head of Protein Sequence Resources, all at EMBL’s European Bioinformatics Institute (EMBL-EBI), explore the research avenues opened up by the AlphaFold database and explain the method's limitations.
: Sameer Velankar and Gerard Kleywegt, from the Protein Data Bank in Europe, and Alex Bateman, Head of Protein Sequence Resources, all at EMBL’s European Bioinformatics Institute (EMBL-EBI), explore the research avenues opened up by the AlphaFold database and explain the method's limitations.
Retrieved from "https://proteopedia.org/AlphaFold"