How to predict structures with AlphaFold

From Proteopedia

Revision as of 21:35, 17 October 2021

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 July, 2021, DeepMind released AlphaFold as open source code. Subsequently, several Colabs became available offering free structure prediction for user-submitted protein sequences. These Google Colabs (collaboratories)^[1]. enable users to submit sequences via web browser, executing the code in the Google cloud, using space private to each user, returning predicted structures.

Below are instructions for beginners who wish to predict structures. We recommend the "advanced" Colab by Sergey Ovchinnikov, Milot Mirdita and Martin Steinegger.

Instructions

Don't worry about any of the options not specifically mentioned below. Leave them at their default settings.
1. Obtain the sequence of the protein of interest, e.g. at UniProt.

2. Login at AlphaFold2_advanced. Registration is free.

3. Paste in your sequence, making sure to completely replace the default sequence:

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.^[2]

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).

5. Scroll down to the subsection titled run alphafold, 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 is often not enough for an optimal result.
• tol (tolerance): Set this to 0.5 (or 1.0 to get a faster result). When a prediction differs from the previous "recycle" prediction by less than this value (RMSD of 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 and num_models. This will proportionally increase the time to complete a result.

6. Open the Runtime menu at the very top of the page, and select Run all.

Downloading Results

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.)

References and Notes

1. ↑ Collaboratory FAQ at Google.
2. ↑ 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.