Nucleic acid residues name truncated in WatsonCrickDist

[mattiabondanza]

Expected behavior

If a_name, c_name etc. arguments are passed as residue names in class initialization, they should be recognized as whole during the analysis, and not assumed to be single-letter.

Actual behavior

Everything except the first letter of the residuename from the groups passed in input is discarded.

Code to reproduce the behavior

from MDAnalysis.analysis import nucleicacids
  wcdist = nucleicacids.WatsonCrickDist(listA, listB, 
                                        a_name='DA', 
                                        c_name='DC',
                                        g_name='DG',
                                        t_name='DT' )
....

My MD for some reasons use this DA DC etc names. for nucleic residues.

Current version of MDAnalysis

MDAnalysis 2.10 on Python 3.13

Proposed solution

In my opinion it should be safe to change resname[0] in resname in lines:

mdanalysis/package/MDAnalysis/analysis/nucleicacids.py

Line 266 in 3be9117

if pair[0].resname[0] in pyrimidines:

mdanalysis/package/MDAnalysis/analysis/nucleicacids.py

Line 268 in 3be9117

elif pair[0].resname[0] in purines:

[BradyAJohnston]

I would definitely say this is a bug and that fix should be all that's required.

Applying this change however would break some existing tests in test_nucleicacids.py.

        g_name: str = "G",
        a_name: str = "A",
        u_name: str = "U",
        t_name: str = "T",
        c_name: str = "C",

The defaults are single letter and the test topologies have 3-letter resnames ('GUA', 'ADE', 'CYT','URA') so would now fail with this internal change. Anyone who is relying on similar functionality would run into issues.

This makes it a bit trickier for how much we want to try and stay backwards compatible.

[mattiabondanza]

Ok I perfectly see your point.
Maybe a compromise solution that keeps backward compatibility while fixing the bug could be something like:

resname_len=max([len(rn) for rn in [a_name, c_name, g_name, t_name, u_name])
...
if pair[0].resname[:resname_len] in purines:

This in the case of all-single letters (that is default parameters) should revert to the previous mechanism, if one needs to specify the names to something different and more than 1 letter long, the correct working mechanism is enforced.

If you think it's worth to completely fix this in a future major a warning could be risen:

if pair[0].resname[:resname_len] in purines and pair[0].resname not in purines:
    # Rise a deprecation warning

Does this fit for you?

[BradyAJohnston]

Might be better that we limit it to just if all residue names are of length 1 then we look at the first letter, otherwise match entire resnames if even one resname is longer than 1.

Even these approaches though I would say are bad because you could match CYT and CYS where one is obviously wrong.

I'm less familiar with this analysis class so I defer to those with more experience with it.