fix dna nucleobase atom selection default for capri

[VGPReys]

Before submitting, read the contributing guidelines and the AI policy.

What does this PR do and why?

This PR modifies the default CG atom selection during caprieval to select the nucelo-bases for DNA (l.316), matching the all-atom selection.

AI assistance

Nope

Checklist

• Tests cover the new and/or changed code
• Documentation updated if needed (also in the haddock3 user-manual
• CHANGELOG.md updated for user-facing changes

Related issues

Closes #1658

Notes for reviewers

Sorry, the automatic linting program did not allow me to touch only the appropriate part, and I had to change the linting all along the script...

[AnnaKravchenko]

Something is not right with definition of U nucleotide (in all-atom):
[2026-08-10 12:01:15,701 libparallel WARNING] Exception in task execution: sequence contains letters not in the alphabet

And then caprieval crushed the entire run

[amjjbonvin]

Do we have any RNA-related tests/examples? If not probably good to add

[AnnaKravchenko]

Technically that is separate issue that is not related to CG atom selection for DNA in capri, but while reading code in this PR, I spotted that in libalign, line 516 (definition of C in RNA_FULL_DICT):

• O2 is missing;
• N9, O6, N2 should not be there

[AnnaKravchenko]

Another potential issue is that DNA_FULL_DICT lists “O1P” and “O2P”, while haddock writes pdbs using “OP1” and “OP2”. I don’t think there’s any way to match those atoms within libalign or caprieval?
So they are probably just discarted when calculating all capri metrics with allatoms=true in caprieval

[AnnaKravchenko]

This is nice, but I initially did not think about the reference file. If that one is not being produced by haddock - then it can, in principle, contain O1P/O2P.

So maybe it’s better to keep all 4 options, same as scan.py in main is doing?

[AnnaKravchenko]

RNA_FULL_DICT[“C”] has “O6” - should be “C6”:

Suggested change

	"O6",
	"C6",

[AnnaKravchenko]

I now copied quotes from the code directly - seems they are now not affected by this suggestion, but indeed beware!

[amjjbonvin]

You are introducing wrong quotes… beware!

[AnnaKravchenko]

One more thing - can you please update caprieval/defaults.yaml of ‘allatoms’ to say that backbone is the case for proteins and RNAs, but it’s nucleobases for DNAs?