Could you provide a minimal example (ideally with some of the examples from the tests, see

I will also say that the GNM code is some of the most ancient analysis code in MDAnalysis and has not seen a lot of maintenance over the years. If you (or someone else) can contribute a way to fix it then that would be very welcome. Otherwise if issues with the code are not being addressed (because developers only have limited amount of time) we may decide to deprecate it and remove in 3.0.

I can't provide a test case because I think there do not have a RIGHT result. However, I can give a demo code to prove that it indeed has a BUG in closeContactGNMAnalysis 's code.

import numpy as np
from MDAnalysis import Universe
from MDAnalysisTests.datafiles import GRO,XTC

u = Universe(GRO,XTC)

gnm = closeContactGNMAnalysis(u, select='name CA', weights='size')
matrix = gnm.generate_kirchoff()
n_residue = matrix.shape[0]
n_accessible_residue = np.array([np.isin(n_residue, res_ids) for res_ids in gnm.ca.residues.ids]).argmax()
print(n_accessible_residue, np.abs(matrix[n_accessible_residue+1:]).sum())

gnm = closeContactGNMAnalysis(u, select='protein', weights='size')
matrix = gnm.generate_kirchoff()
print(np.abs(matrix[n_accessible_residue+1:]).sum())

It will output:

14 0.0
6894.190527601791

Where 14 is the max residue the closeContactGNMAnalysis can use with select 'name CA', so that you can see that, after residue 14, the matrix ONLY has ZERO score. However, it do not make sense as the 'protein' selection has 6894.190527601791 score after residue 14.

If you want to make a fix, I can give my fix code.

Thanks for contributing a fix in PR #4961 !