How to predict structures with AlphaFold: Difference between revisions

From Proteopedia
Jump to navigationJump to search
← Older edit
Eric Martz (talk | contribs)
ConfirmAccount, Editor, Print3D, Tester, journal, Administrators
30,118 edits
No edit summary
Eric Martz (talk | contribs)
ConfirmAccount, Editor, Print3D, Tester, journal, Administrators
30,118 edits
No edit summary
 
(162 intermediate revisions by the same user not shown)
Line 1: Line 1:
In 2020, the AlphaFold project of Google's DeepMind team demonstrated a major breakthrough in predicting protein structure from sequence. Their success in the blind CASP competition astonished many experts. For an overview, see [[Theoretical models]].
In 2020, the [[AlphaFold]] project of Google's DeepMind team demonstrated a major breakthrough in predicting protein structure from sequence. Their success in the blind [[CASP]] competition astonished many experts. For an overview, see [[Theoretical models]], bearing in mind "The Joys and Perils of AlphaFold"<ref name="perils">PMID: 34668287</ref>. AlphaFold2 continued to have the highest success rate in the [[Theoretical_models#2022:_CASP_15|2022 CASP 15 competition]]. In 2024, the AlphaFold team won half of the [[Nobel_Prizes_for_3D_Molecular_Structure#2020-2029|Nobel Prize in Chemistry]].


In July, 2021, DeepMind released [[Alphafold#AlphaFold_published_July_2021|AlphaFold as open source code]]. Subsequently, [[Alphafold#Free_AlphaFold-based_Servers|several Colabs became available]] offering '''free''' structure prediction for user-submitted protein sequences. These Google Colabs (collaboratories)<ref name="colabfaq">[https://research.google.com/colaboratory/faq.html Collaboratory FAQ] at Google.</ref>. enable users to submit sequences via web browser, executing the code in the Google cloud, using space private to each user, returning predicted structures.
In July, 2021, DeepMind released [[Alphafold#AlphaFold_published_July_2021|AlphaFold as open source code]]. Subsequently, [[Alphafold#Free_AlphaFold-based_Servers|several Colabs became available]] offering '''free''' structure prediction for user-submitted protein sequences. These Google Colabs (collaboratories)<ref name="colabfaq">[https://research.google.com/colaboratory/faq.html Collaboratory FAQ] at Google.</ref>. enable users to submit sequences via web browser, executing the code in the Google cloud, using space private to each user, returning predicted structures. In 2024, DeepMind provided the [https://alphafoldserver.com AlphaFold3 server]<ref name="af3">PMID: 38718835</ref> (see below).


Below are instructions for beginners who wish to predict structures.
Below are instructions for beginners who wish to predict structures.
==Is An Empirical Model Available?==
Empirical models are the most accurate, so you should look for those first. See [[How To Find A Structure]]. If there is no empirical model for your amino acid sequence, it may be useful to explore [[How_To_Find_A_Structure#Sequence-Related_Empirical_Models|empirical models for closely-related sequences]], if available. Even if an empirical structure is available, most have missing residues or atoms, and it may be useful to compare it with the AlphaFold prediction: see [[Missing residues and incomplete sidechains]].
==Does AlphaFold Database Already Have Your Protein?==
Structure predictions for over 200,000,000 proteins are available from the [https://alphafold.ebi.ac.uk/ AlphaFold Database]. If your protein is there, download the prediction from the Database. Then you can explore it in the viewer of your choice. For beginners, [http://firstglance.jmol.org FirstGlance in Jmol] is easiest if you want more than a momentary impression. [http://firstglance.jmol.org/where.htm#uploading Upload] your PDB file. FirstGlance has numerous [http://firstglance.jmol.org/whatis.htm#unique unique conveniences] yet considerable depth and power.
In 2024, AlphaFold Database predictions are always [https://alphafoldserver.com/faq single protein chain structures without ligands]. If your protein is an assembly of multiple chains, you will likely want to compare the Database structure with predictions from the latest servers capable of multiple-chain + ligand predictions (see below).
==Prediction Servers==
You can submit one (or a set) of sequences to these servers, and they will return predicted structures, along with estimates of confidence in their predictions. This is not a comprehensive list. Please add other servers of interest to a broad range of users, including beginners.
* 2024<ref name="af3" />: [https://alphafoldserver.com AlphaFoldServer.Com]. Using AlphaFold3, predicts homo- and hetro-multimers involving protein, DNA, RNA, ligands, and modified residues. Straightforward to use; Guide and FAQ provided. Predictions are templated without user control (see FAQ). Free for non-commercial use -- see [https://alphafoldserver.com/terms Terms of Service] and [https://alphafoldserver.com/output-terms Output Terms of Use]. From the DeepMind team<ref name="af3" />.
**If you get <font color="red"><b>Invalid character</b></font> after pasting in your sequence, try removing the breaks between lines.
**Predicted models are in [[Mmcif format|mmCIF format]] only. When visualized in [http://firstglance.jmol.org FirstGlance in Jmol], coloring by confidence/pLDDT is automatic. [[Converting AlphaFold3 CIF to PDB|More information about .cif and FirstGlance]].
**To easily obtain average [[pLDDT]] (predicted confidence) for a range of residues, see [[FirstGlance/How to get average pLDDT from AlphaFold models]].
**Chain ID assignments differ from those of empirical wwPDB files: see [[Chains_and_Chain_IDs#AlphaFold3_Chain_IDs|AlphaFold3 Chain IDs]].
**See also [[#Visualizing Predicted Structures]] and [[User:Eric Martz/AlphaFold3 case studies|AlphaFold3 case studies]].
* 2024<ref name="rfaa">PMID: 38452047</ref>: RosettaFold All-Atom (RFAA) predicts multimers of protein and nucleic acids with ligands. From the Baker team<ref name="rfaa" />. [https://neurosnap.ai/service/RoseTTAFold%20All-Atom A free server limited to very small numbers of jobs is available from Neurosnap].
* 2024 update? [https://colab.research.google.com/github/sokrypton/ColabFold/blob/main/AlphaFold2.ipynb ColabFold AlphaFold2 and AlphaFold-multimer] predicts multiple chain complexes. Easy to use. From Mirdita ''et al.''<ref name="colabfold" />
* 2024<ref name="combfold">PMID: 38326495</ref>: [https://neurosnap.ai/service/CombFold CombFold] predicts the structures of large protein complexes from subunit sequences using AlphaFold Multimer paired with a combinatorial method to assemble subunits. From Shor and Schneidman-Duhovny<ref name="combfold" />.
* 2022<ref name="colabfold">PMID: 35637307</ref>: [https://colab.research.google.com/github/sokrypton/ColabFold/blob/main/beta/AlphaFold2_advanced.ipynb ColabFold AlphaFold2_advanced]. Predicts homo- and hetero-multimers using methods from the Steinegger/Mirdita team<ref name="colabfold" /><ref name="colabfold2024">PMID: 39402428</ref>, before AlphaFold-multimer<ref name="afmultimer">[https://www.biorxiv.org/content/10.1101/2021.10.04.463034v1.full Protein complex prediction with AlphaFold-Multimer], Preprint, Evans et al. 2021.</ref> was available. Does NOT use templates. [https://www.ebi.ac.uk/training/online/courses/alphafold/accessing-and-predicting-protein-structures-with-alphafold/predicting-protein-structures-with-colabfold-and-alphafold-colab/ See Instructions from EMBL-EBI].
*2022<ref name="af2021">PMID: 34265844</ref><ref name="afmultimer" />: [https://colab.research.google.com/github/deepmind/alphafold/blob/main/notebooks/AlphaFold.ipynb AlphaFold2/Multimer Colab] able to predict protein multimers. From the DeepMind team<ref name="af2021" /><ref name="afmultimer" />. [https://www.ebi.ac.uk/training/online/courses/alphafold/accessing-and-predicting-protein-structures-with-alphafold/predicting-protein-structures-with-colabfold-and-alphafold-colab/ See Instructions from EMBL-EBI].
*2022<ref name="alphafill">PMID: 36424442</ref>: [https://alphafill.eu/ AlphaFill] “transplants” missing ligands, cofactors and (metal) ions into AlphaFold models. From the Perrakis team<ref name="alphafill" />. Ligand positioning is approximate. See [[Alphafold#Ligands:_AlphaFill |CAUTION]] provided by the AlphaFill team:
<blockquote>
"AlphaFill models are not meant or suitable for precise quantification of interactions between the transferred ligand(s) and the protein (e.g. hydrogen bonds, π-π or cation-π interactions, van der Waals interactions, hydrophobic interactions, halogen bonds)."
</blockquote>
*2021<ref name="rosettafold">PMID: 34282049</ref>: [https://robetta.bakerlab.org/ RoseTTAFold at Robetta] is an independent design from the Baker team<ref name="rosettafold" />, influenced by the design of AlphaFold2. Predicts monomers and multimers. Comparing results of RoseTTAFold with results of AlphaFold2/3 is worthwhile. At [https://robetta.bakerlab.org/ Robetta], open the Structure Prediction menu at the top, and choose Submit. ''Be sure to check RoseTTAFold under Optional!''
===Cost?===
The above servers are free for limited, non-commercial use.
'''Colabs''': After multiple free jobs in a Colab, a new job may be refused. You may be informed that a GPU could not be assigned. In 2024, a subscription to [https://colab.research.google.com/signup/pricing Colab Pro] is US $10/month. Paying this will enable you to do many more jobs.
==Visualizing Predicted Structures==
[http://firstglance.jmol.org FirstGlance in Jmol] automatically colors its initial view of AlphaFold or RoseTTAFold models by estimated confidence [[pLDDT]] ('''{{Font color|blue|blue for high confidence}}, {{Font color|red|red for low confidence}}'''). After you go to other views or tools, you can always get back to this color scheme by clicking ''Reliability Estimates'' in the ''Views'' tab.
*[[iCn3D]] automatically colors AlphaFold2 Database models loaded from their UniProt IDs. For AlphaFold files opened from your computer, use pLDDT on the pull-down Color menu.
*[[PyMOL]] and [[ChimeraX]] have no built-in confidence/pLDDT color scheme. Their rainbow/spectrum color schemes for temperature/B-factor color confidence/pLDDT with the AlphaFold color scheme inverted.
[http://firstglance.jmol.org/where.htm#uploading Upload] your predicted PDB file to [http://firstglance.jmol.org FirstGlance.Jmol.Org], which has many [http://firstglance.jmol.org/whatis.htm#unique unique conveniences and capabilities].
When using [http://FirstGlance.jmol.org FirstGlance], it is easy to visualize
* Estimated confidence/pLDDT by touching an atom
* '''Average confidence/pLDDT''' ("reliability") for the entire model, or for [[FirstGlance/How to get average pLDDT from AlphaFold models|a specified sequence range]].
* Secondary structure (Views tab)
* Distribution of hydrophobic vs. polar residues (Views tab: integral membrane proteins will have large hydrophobic surfaces while soluble proteins will have hydrophobic cores revealed by the ''Slab'' button)
* Distribution of charges (Views tab: nucleic acid binding sites will have clusters of positive charges)
* Disulfide bonds (Tools tab)
* Domain structure and positions of the ends of the polypeptide chain (Views tab: N -> C Rainbow)
* Locations of functional sites by evolutionary conservation (see instructions at [[How_to_see_conserved_regions]])
==Instructions for ColabFold 2022==
This procedure was written in 2022<ref name="">Some of the recommended options were gleaned from the [https://www.youtube.com/watch?v=Rfw7thgGTwI 1 hour 46 minute video of presentations by Sergey Ovchinnikov and Martin Steinegger] (August, 2021) for the Boston Protein Design and Modeling Club
hosted by Chris Bahl.
</ref>. In 2024, ColabFold is not necessarily the best or only place to submit you job: see [[#Prediction Servers]].
Initially, AlphaFold and ColabFold performed best with '''single chains'''<ref name="afmultimer">[https://www.biorxiv.org/content/10.1101/2021.10.04.463034v1 Protein complex prediction with AlphaFold-Multimer], 2021, Evans ''et al.'' (DeepMind Team).</ref>, which may include one or a few domains. The instructions below were written '''before ColabFold was adapted to prediction of multimers'''. If you are interested in complexes or alternate conformations, please see ColabFold instructions in the 2023 paper by Kim ''et al.'' <ref name="kim2023">[https://protocolexchange.researchsquare.com/article/pex-2490/v1 Easy and accurate protein structure prediction using ColabFold], 2023, Kim ''et al.'' (DeepMind Team).</ref>
===Submitting A Sequence===
First, if your query is a single chain molecule, check the [https://alphafold.ebi.ac.uk/ AlphaFold Database] for the protein of interest. If its structure has already been predicted there, download it, and skip to [[#Interpreting Results|Interpreting Results]] below. Otherwise ...
Don't worry about any of the options not specifically mentioned below. Leave them at their default settings.
<br>
1. Obtain the sequence of the protein of interest, e.g. at [http://uniprot.org UniProt]. Click on the FASTA button above the sequence in UniProt. Copy only the sequence, excluding the FASTA header line that begins with ">".
<br><br>
2. Login with a google account at [https://colab.research.google.com/github/sokrypton/ColabFold/blob/main/beta/AlphaFold2_advanced.ipynb AlphaFold2_advanced]. You can register for a free gmail account to use for login.
(Another free AlphaFold2 service is [https://colab.research.google.com/github/sokrypton/ColabFold/blob/main/AlphaFold2.ipynb ColabFold]. Using it may require a procedure different from the steps below.)
<br><br>
3. Paste in your sequence, making sure to completely replace the default sequence:
<br>
[[Image:AF2Adv-seq1.png|400px]]
<br>
This input slot can accept sequences >1,000 amino acids, even though it is only one line. Sequence lengths of ~1,000 amino acids, or longer, may cause the Colab to fail, but can be predicted by submitting in two halves.<ref name="halves">I had one sequence of length ~1,300. After it failed, I submitted it as two halves with a substantial overlap (~350 residues). The middle overlap of ~200 residues of the predicted structures superposed very closely with [[DeepView]]. I trimmed off the ends that superposed poorly, and superposed the two halves via the mid-overlap. By inspection, I chose pair of alpha carbons near the middle where the alpha carbon positions were nearly identical. I trimmed each half to this position, and "ligated" the two halves by combining the superposed half PDB files with a text editor. For further details, contact [[User:Eric_Martz]].</ref> See also <ref name="halves" /> and [[Joining AlphaFold predictions for halves of a molecule]].
<br><br>
4. Enter a jobname in the slot below the sequence slot. The results.zip filename will begin with this jobname (but none of its contents include the jobname).
<br><br>
5. Scroll down to the section titled ''run alphafold'', subsection ''Sampling options'':
* '''num_models''', the number of models to be predicted, is 5 by default. You could reduce this to 3 if you are in a hurry.
* '''max_recycles''': Set this to 48 (or at least 12). The actual number of "recycles" performed will stop when the model has converged to the specified tolerance. The default of 3 recycles is often not enough for an optimal result.
* '''tol''' (tolerance): Set this to 0.5 &Aring; (or 1.0 to get a faster result). When a prediction differs from the previous "recycle" prediction by less than this value (RMSD in &Aring; between alpha carbons), the recycles will stop.
* '''num_samples''' (random seeds): Leave this at 1. Beware that if you increase this above 1, you will generate a number of models equal to the product of this value times num_models. This will proportionally increase the time to complete a result.
<br>
6. Open the Runtime menu at the very top of the page, and select '''Run all'''.
<br>
[[Image:AF2Adv-runall.png]]
<br>
Don't worry about the "Warning". It is just Google's disclaimer that they did not write the code you are about to execute. Click ''Run anyway''.
===Downloading Results===
{{Font color|red|Do NOT close your AlphaFold2_advanced browser tab until the job is completed.}} It appears that you will lose your job if you close the browser tab. You will be warned if you inadvertently try.
When the job is completed, a dialog to download a zip file will appear automatically. (Sometimes you will be asked for permission to enable download first.)
===Interpreting Results===
Static images of backbone renderings of predicted models will appear in your web browser at the bottom of the section ''run alphafold'' as each is completed.
====Estimated Reliability====
Each predicted model has an average estimated reliability ([[AlphaFold pLDDT and expected distance error|pLDDT]], predicted local distance difference test). >90 is likely accurate; <70 is low confidence. For more about interpreting these values, please see the [https://alphafold.ebi.ac.uk/faq AlphaFold Database FAQ].
Each residue has an estimated reliability of its position (0-100) in the PDB [[temperature]] column. BEWARE that high values mean high confidence, and low values mean low confidence. This is the INVERSE of [[temperature|crystallographic temperature values]], where low values are good and high values are bad. Uploading your PDB file to [http://firstglance.jmol.org FirstGlance in Jmol] will automatically color each residue by its estimated reliability.
====Intrinsic Disorder====
Some models have high confidence in a folded [[domain]], and low confidence in a segment that is not part of a compact domain. Low-confidence segments may be [[Intrinsically_Disordered_Protein|intrinsically disordered]]. It is useful to compare [[Intrinsically_Disordered_Protein#Protein_disorder_predictors|predictions of disorder]] with AlphaFold reliability estimates.
====Relative Positions of Domains====
If the predicted model has more than one [[domain]], each domain may have high confidence, yet the relative positions of the domains may not. The estimated reliability of relative domain positions is in graphs of '''predicted aligned error''' (PAE) which are included in the downloadable zip file of results. For an explanation, see ''How should I interpret the relative positions of domains?'' in the [https://alphafold.ebi.ac.uk/faq AlphaFold Database FAQ].
====Recycles For Convergence====
You may be interested to note the number of recycles required for each model to converge to the specified tolerance. These numbers are not captured in the downloaded zip file.
The models will be ranked with number one having the highest estimated reliability ([[AlphaFold pLDDT and expected distance error|pLDDT]]). This is usually not in the order in which they were calculated. You might want to copy the ranking list, perhaps adding the number of recycles and final tolerance values:
<pre>
model rank based on pLDDT              Recycles  Tolerance
rank_1_model_2_ptm_seed_0 pLDDT:62.46    10          0.33
rank_2_model_3_ptm_seed_0 pLDDT:59.59    9          0.47
rank_3_model_1_ptm_seed_0 pLDDT:55.63    12          0.52
</pre>
Notice that the model predicted 2nd had the best estimated reliability (pLDDT), and that the model ranked 3rd did not quite achieve the specified tolerance of 0.5 &Aring; RMSD after 12 recycles. (12 was specified as the maximum in this job.)
Also notice that, in this case, all 3 models have low confidence (pLDDT < 70), and are of questionable value.
==See Also==
*[[AlphaFold/Index]], a list of pages in Proteopedia about Alphafold.
*[[User:Eric Martz/AlphaFold3 case studies|AlphaFold3 case studies]] includes a case that AlphaFold3 cannot predict.
*[[How To Find A Structure]]
*[[Missing residues and incomplete sidechains]]


==References and Notes==
==References and Notes==
<references />
<references />
Retrieved from "https://proteopedia.org/How_to_predict_structures_with_AlphaFold"