Question about indices from FastNS.get_pairs()

[erwusht]

Hi,

I'm working with MDAnalysis and have a question about the indices returned by the FastNS.search (NSResults.get_pairs()) method.

Here's the relevant code I'm using:

import MDAnalysis as mda
from MDAnalysis.lib.nsgrid import FastNS

u = mda.Universe('protein.pdb', 'md_center.xtc')
protein_atoms = u.select_atoms("protein")

pair_counts = np.zeros((len(protein_atoms), len(protein_atoms)))

groupA = u.select_atoms('resid 1-20 and not type H')
groupB = u.select_atoms('resid 20-30 and not type H')

ns = FastNS(cutoff=4.8, coords=groupA.positions, box=u.dimensions)
ns.search(groupB.positions)
pairs = ns.get_pairs()

for i, j in pairs:
    # Is it right for j representing groupA and i representing groupB?
    pair_counts[groupA[j].index, groupB[i].index] += 1

My understanding is that i represents the index of an atom from groupB, and j represents the index of an atom from groupA. Is this correct?

Any help would be greatly appreciated!

Thank you!
Shiauthie Wu

[orbeckst]

My understanding is that i represents the index of an atom from groupB, and j represents the index of an atom from groupA. Is this correct?

Based on the docs https://docs.mdanalysis.org/stable/documentation_pages/lib/nsgrid.html#MDAnalysis.lib.nsgrid.FastNS.search that would be my interpretation.

@hmacdope @richardjgowers does this sound right?

[orbeckst]

How would you rephrase the docs to make them clearer?

@hmacdope @richardjgowers do you have an opinion?

[hmacdope]

Sorry for the slow reply, so my understanding is that we use two sets of nomenclature, where we probably could just one

M understanding is that

query[i] == reference[i]
coordinates[j] == configuration[j]

If this is correct (@richardjgowers?), I think we should probably go with changing to reference and configuration IMO as these make the most logical sense (to me).

reference

[orbeckst]

@erwusht could you open an issue to improve the docs for FastNS? Following @hmacdope 's assessment, we should use consistent terminology.

Please summarize the problem, suggest the change, and link to this discussion.

Once we have an issue, we know that this is something that should get fixed and then maybe you or another volunteer can make the changes. If it's not an issue then it will be forgotten.

Thank you!

[erwusht]

@orbeckst @hmacdope Sorry for the delayed response.

Thank you very much for being willing to take the time to solve this problem.

I'm willing to contribute towards this improvement of the docs. I'm also happy to offer some ideas as references for the rephrased versions.

Maybe a few days later, I need some time to do this.

[orbeckst]

Any contributions are very welcome! I suggest starting with opening an issue because that's the point where we can then discuss what needs to be done. You can then just open a PR that references the issue.