Calculate specific atom pair distances with MDTraj

atom_pair_distances.py

import itertools
import mdtraj, mdtraj.geometry
import pandas as pd

traj = mdtraj.load("./traj.xtc", top="native.pdb")

top, bonds = traj.top.to_dataframe()

atoms = np.array(["H", "HA", "N", "CA", "C", "CB"])

bad_residues = np.array(["GLY", "PRO"])

top = top[np.in1d(top.name, atoms)]  # Selected desired atoms
top = top[~np.in1d(top.resName, bad_residues)]  # Remove unwanted residues

atom_pairs = []
for k, x in top.groupby("resSeq"):
    atom_pairs.extend(list(itertools.combinations(x.index, 2)))

atom_pairs = np.array(atom_pairs)

distances = mdtraj.geometry.distance.compute_distances(traj, atom_pairs)


id0 = top.ix[atom_pairs[:,0]].resSeq
id1 = top.ix[atom_pairs[:,1]].resSeq
name0 = top.ix[atom_pairs[:,0]].name
name1 = top.ix[atom_pairs[:,1]].name

columns = pd.MultiIndex.from_arrays([id0, name0, id1, name1])
distances = pd.DataFrame(distances, columns=columns)

@tjlane, @rmcgibbo, @osamae

Here is the preliminary code to load a trajectory, find the desired pairs of atom indices, and calculate the distances.

PS Osama, some of those distances are going to be constant during the MD simulation, as we typically constrain the bond lengths. If you like, you can try to make a better script that deletes those atompairs.

If speed is an issue and you have a lot of frames of MD, you can also use msmbuilder.geometry.contact.atom_distances, which is the same function as mdtraj.geometry.distance.compute_distance, but probably a little faster. I haven't ported that C code to mdtraj yet.

And this requires mdtraj version 0.4.0, which was just released yesterday. you should be able to upgrade with pip install mdtraj --upgrade if you haven't already.

PS: Robert it's super convenient having access to the element column, which I never had in any previous PDB readers...

I added an optional conversion of the distance matrix into DataFrame...