PfATP4 structure predicted with AlphaFold2

[miquelduranfrigola]

Hi @edwintse @mattodd

Last week the code of AlphaFold2 was published. Today EBI and DeepMind are releasing full structural proteomes predicted with this AI tool: https://alphafold.ebi.ac.uk/. Also InterPro and PFam now provide AlphaFold2 predictions. All very exciting.

I don't think PfATP4 (Q9U445) structure prediction is available yet, so we have run AlphaFold2 ourselves. Perhaps this is useful to revisit the docking approaches :)

• Please see this repository containing PDB files for 3 model runs of the algorithm: https://github.com/ersilia-os/osm-pfatp4-structure
• We have prepared a Google Colab notebook that you can run to reproduce results. If you scroll down you will see a 3D display of the model.

Note that we truncated the first 115 residues because they failed to align and eventually fold. HHblits was used for the MSA alignment.

[mattodd]

That's really super, thanks @miquelduranfrigola! Pinging @holeung and will mail Kiaran Kirk. Indeed, the thing I'd like to do here is trial, what, 5 known actives and 5 known inactives and try to build on the picture that Ho Leung constructed as to what the relevant binding interactions are, described in outline in the current version of the Series 4 mail paper that is nearing completion:
http://tinyurl.com/OSM-S4-Paper1

[holeung]

Yes! Like everyone else in our research community, we can't stop talking about AlphaFold. Will take a look and get back to everyone.

[holeung]

One of our PhD students, Ye Zou, took a close look at the AlphaFold2 predictions for PfATP4. She found it generally performed worse than traditional homology models. We are currently working on new homology models that can take advantage of PDB 6YAA, a crystal structure of sarco/endoplasmic reticulum calcium ATPase from rabbit, that was solved in 2020. This structure is special because it contains a bound inhibitor. The inhibitor binds in the same binding pocket I predicted earlier from my older homology models.
YeZou_comparison_summary.pdf

Ye Zou's full analysis is attached.
YeZou Presentation1.pptx

[mattodd]

Really nice @holeung, thanks for the update and it'll be interesting to see the new model. The inhibitor you mention in 6YAA. Is it active vs plasmodium...?

[holeung]

The inhibitor in 6yaa is thapsigargin, which I believe has no/weak activity against plasmodium.

[tewarisg]

One of our PhD students, Ye Zou, took a close look at the AlphaFold2 predictions for PfATP4. She found it generally performed worse than traditional homology models. We are currently working on new homology models that can take advantage of PDB 6YAA, a crystal structure of sarco/endoplasmic reticulum calcium ATPase from rabbit, that was solved in 2020. This structure is special because it contains a bound inhibitor. The inhibitor binds in the same binding pocket I predicted earlier from my older homology models. YeZou_comparison_summary.pdf

Ye Zou's full analysis is attached. YeZou Presentation1.pptx

Hi @holeung, we would appreciate it if you could share the homology model of PfATP4 developed by your student. Is that available online? Another thing that we noticed in these AlphaFold models is incorrect location of several amino acids, e.g., Glycine is present at the 211 position instead of Alanine. Thanks!

[holeung]

Ye Zou's AF2 model is attached.

PFATP4_3temp_116_end_relaxed_model_1.pdb.zip

[miquelduranfrigola]

One of our PhD students, Ye Zou, took a close look at the AlphaFold2 predictions for PfATP4. She found it generally performed worse than traditional homology models. We are currently working on new homology models that can take advantage of PDB 6YAA, a crystal structure of sarco/endoplasmic reticulum calcium ATPase from rabbit, that was solved in 2020. This structure is special because it contains a bound inhibitor. The inhibitor binds in the same binding pocket I predicted earlier from my older homology models. YeZou_comparison_summary.pdf
Ye Zou's full analysis is attached. YeZou Presentation1.pptx

Hi @holeung, we would appreciate it if you could share the homology model of PfATP4 developed by your student. Is that available online? Another thing that we noticed in these AlphaFold models is incorrect location of several amino acids, e.g., Glycine is present at the 211 position instead of Alanine. Thanks!

Thanks for this! I used the UniProt AC: Q9U445. Is this the one we should be using? It is an unreviewed entry in UniProt...

[tewarisg]

Thanks for the reply. Yes, I did first review the Uniprot sequence that you used to predict the structure. The Uniprot sequence is correct, e.g., Alanine at 211, Glycine at 223, etc. Mutations at these positions are known for conferring resistance against spiroindolone and pyrazoleamide compounds. For some reason, the AlphaFold2-predicted structures have GLY and LEU at these positions. Do you know why?

[tewarisg]

Ye Zou's AF2 model is attached.

PFATP4_3temp_116_end_relaxed_model_1.pdb.zip

Got it. Thanks for the quick reply. It appears that this predicted model also has GLY at 211 and LEU at 223 position. As per UniProt AC: Q9U445, the model should have ALA at 211 and GLY at 223 position. Do you know why the predicted model has different amino acids at these locations?

[miquelduranfrigola]

Hi @tewarisg - I will have a deeper look as soon as I can. If you check the repo: https://github.com/ersilia-os/osm-pfatp4-structure you will see that we trimmed the first 115 residues of the sequence, as they were not folding successfully. Perhaps this is the reason why you see a mismatch in the sequence positions?

[tewarisg]

Hi @tewarisg - I will have a deeper look as soon as I can. If you check the repo: https://github.com/ersilia-os/osm-pfatp4-structure you will see that we trimmed the first 115 residues of the sequence, as they were not folding successfully. Perhaps this is the reason why you see a mismatch in the sequence positions?

Hi @miquelduranfrigola, You are correct. I am able to find all the important residues after subtracting 115 from their original positions. Thanks!!