Error while fetching from AlphaFold database

[m.bienstein@…]

Dear ChimeraX Team,

First of all, thank you for this really satisfying program.
When I was using the AlphaFold tool implemented in ChimeraX (latest version), which is really nice by the way, I notices a small error, which probably can be fixed easily:
While searching similar proteins (BLAST) that I wanted to fetch out of the result list with the AlphaFold tool, the following error appeared:
"Fetching url https://alphafold.ebi.ac.uk/files/AF-O08337-F1-model_v1.cif failed:
HTTP Error 404: Not Found"
I would like to highlight here that I am not speaking about the fetching itself but about the fetching after searching.

I was curious and searched the model manually on the database. I could find it, when I just increases the version number of the model to v2.

Finally, the following reference worked well:
https://alphafold.ebi.ac.uk/files/AF-O08337-F1-model_v2.cif

Maybe this issue was already raised and will be included in the next version, but otherwise I hope I could help to improve this nice program a little bit more.


Best regards,

Marian Bienstein



Marian Bienstein (M. Sc.)
ABBt Team Schwaneberg
Department of Biotechnology
RWTH Aachen University
Phone: +49 241 80 20660
Mail: m.bienstein@biotec.rwth-aachen.de<mailto:m.bienstein@biotec.rwth-aachen.de>

[Tom Goddard]

Thanks for the info. Were you using ChimeraX 1.3 when you saw this problem? (You can use menu Help / Report a Bug... to send me your ChimeraX version info.) When ChimeraX 1.3 came out only the v1 version of the EBI AlphaFold database was out. When the introduced the v2 version in January 2022 the ChimeraX daily build was updated to use that but ChimeraX 1.3 that was already distributed still used v1. It was working in the past since EBI left the v1 AlphaFold models online. I'll have to check if they have removed those files. That would be unfortunate. At any rate, you can use the ChimeraX daily build to use the v2 files.

EBI also updated the the AlphaFold database a second time in late January or February. The did not call it version v3, and instead just added 200,000 entries to v2. ChimeraX search is not yet finding those 3 revision files because I have not yet updated the AlphaFold sequence database on our servers that run the ChimeraX BLAST search. I plan to do that soon. We have been busy the past couple months trying to renew ChimeraX funding, so did not have time to keep this up to date.

[Tom Goddard]

Just tested ChimeraX 1.3 AlphaFold fetch and search with pdb 7q50 (human protein). It worked correctly fetching

​https://alphafold.ebi.ac.uk/files/AF-Q8IVV7-F1-model_v1.cif

Looking at our AlphaFold sequence database, it does not have your model UniProt O08337. But when I use that UniProt ID with the AlphaFold GUI search in ChimeraX 1.3 it lists that sequence as the top hit. That should not happen. The ChimeraX 1.3 BLAST search should only be searching the AlphaFold v1 sequences, but apparently it is search the v2 sequences.

[Tom Goddard]

Zach, can you look into how ChimeraX 1.3 is searching the AlphaFold v2 database. It has to search the v1 database because the client code only knows how to fetch v1 database entries.

[Zach Pearson]

Yes, this is definitely an error. Because ChimeraX 1.3 does not even know it needs to specify a version, we assume a value on the backend when we see a request come in without a version. I haven't looked yet, but my feeling is that I set it to '2' or whatever the latest version is by default, when it should be '1'.

I will make that change on the backend to fix ChimeraX 1.3.

[goddard@…]

Sounds good.  When you make the change, try searching AlphaFold for the sequence of uniprot O08337 in  ChimeraX 1.3 and make sure the first hit in the result loads and is not O08337 (since it is only in v2).  Then do the same test in ChimeraX daily build where the first hit should be O08337 and should load from v2 database.

[Zach Pearson]

Switched the default version on the backend from 2 to 1 and ran those tests. Should be working normally now!