Another issue I've noticed is that for objects which have residue sequence data available, but the residues don't have structural information (like in loops), this script is unable to find the sequence to select.

But I'm not sure how to update the selection behavior to fix it.

I'll check the firstOnly issue soon. Can you dump here some script that check findseq use case?

And about your missing residue issue, I have an idea that seems to work. For sure the sequence data is available in RCSB PDB and mmCIF files but may be missing when they are from other sources, I don't know when it is the case.

The API only cmd.get_fastastr is a command related to findseq, may you see if it works in your case? In my case the FASTA string is retrieved complete but iterate and cmd.get_model aren't, and I don't know why.

https://github.com/schrodinger/pymol-open-source/blob/9d3061ca58d8b69d7dad74a68fc13fe81af0ff8e/modules/pymol/exporting.py#L169

The ONE_LETTER table has some errors like the map 'CRF':'TWG' which will ruin the analysis, in case of matching. There are also cases like 'A ':'A', which are not useful.

Edit: I checked some values in ONE_LETTER and I don't trust it.

I reverted your commits as I tested it a few hours ago and they weren't working. I shouldn't had merged it without testing.

You can recover your commits by this PR branch, if you need. Take your time...

cmd.get_fastastr gave the wrong fasta string for my test file (1qys), returning a sequence with "?" instead of "M". However, cmd.iterate was able to retrieve those residues without structural information, so now we can find them as well. I took the liberty of renaming the firstOnly to matchMode to reflect the behavior for multiple objects.

It worked as expected on my files, but I'd need your files and tests to check if it is working as expected.

Note that the residues without structural information can be retrieved only if using a file with appropriate metadata. In custom made files it will skip missing residues, I guess. If pertinent, it should be handled by code or explicitly stated at documentation.

Note that the residues without structural information can be retrieved only if using a file with appropriate metadata. In custom made files it will skip missing residues, I guess. If pertinent, it should be handled by code or explicitly stated at documentation.

Do you have one of these so I can test the code?

I tested your branch patch-1, and it's working but conflicting. Can you force push or something?

I don't know how to force push, so I edited the conflict thing.

Another thing that I don't know if we should document on the help function is that this code (and the original version as well) only search for non-overlapping matches. In a future update, how about we add the option to create a group and put each match into its own selection, which would allow for overlapping matches?

I don't understand this regex in the examples. It is multi-character? How it works?

        # find the Potential N-linked glycosylation sites in 5fyj
        fetch 5fyj
        findseq N(?=[^P][ST]), 5fyj and chain G+B, 5fyj_pngs

I don't understand this regex in the examples. It is multi-character? How it works?

        # find the Potential N-linked glycosylation sites in 5fyj
        fetch 5fyj
        findseq N(?=[^P][ST]), 5fyj and chain G+B, 5fyj_pngs

It looks for a single character. This regex is using a lookahead assertion to match only N's that are followed by XY where X is not P and Y can be either S or T.

From wikipedia:

On the other hand, the attachment of a glycan residue to a protein requires the recognition of a consensus sequence. N-linked glycans are almost always attached to the nitrogen atom of an asparagine (Asn) side chain that is present as a part of Asn–X–Ser/Thr consensus sequence, where X is any amino acid except proline (Pro).[4]