Created
October 18, 2019 23:17
-
-
Save JoaoRodrigues/ce56e74cc527b63c429d46b707da7846 to your computer and use it in GitHub Desktop.
Assorted Python function for plotting molecular properties
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
# Copyright 2019 João Pedro Rodrigues | |
# | |
# Licensed under the Apache License, Version 2.0 (the "License"); | |
# you may not use this file except in compliance with the License. | |
# You may obtain a copy of the License at | |
# | |
# http://www.apache.org/licenses/LICENSE-2.0 | |
# | |
# Unless required by applicable law or agreed to in writing, software | |
# distributed under the License is distributed on an "AS IS" BASIS, | |
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | |
# See the License for the specific language governing permissions and | |
# limitations under the License. | |
# Imports required for the code to work | |
import collections | |
import matplotlib.pyplot as plt | |
import numpy as np | |
import mdtraj as md | |
# Assorted functions with proper documentation. | |
# | |
# == HBond Viewpoint Explorer == | |
# Plots 'barcode' plots of all residues hydrogen bonding | |
# with a specific viewpoint residue. | |
# | |
def hbond_viewpoint(trajectory, residue, intramolecular=False): | |
"""Produces a 2D numpy array with hydrogen bonding data. | |
trajectory: mdtraj Trajectory object | |
residue: mdtraj Residue object | |
Returns a ResxFrame 2D numpy binary array with info on | |
when two residues are hydrogen bonding over a trajectory. | |
Also returns a dictionary with row idx (1st Dim) and mdtraj | |
Residue object, for labeling purposes. | |
""" | |
t = trajectory | |
# 1. Calculate contacts of residue (ca only) | |
ca_d_max = 12.0 | |
atomsel = t.top.select("name CA") | |
residsel = {a.index for a in residue.atoms} & set(atomsel) | |
t_nl = md.compute_neighbors(trajectory, ca_d_max, | |
haystack_indices=atomsel, query_indices=residsel) | |
# 2a. Obtain unique set of atom ids in contact with the viewpoint | |
atom_nl = {a for f in t_nl for a in f} | |
# 2b. Expand selection to residues | |
res_nl = [t.top.atom(a).residue for a in atom_nl] | |
# 2c. Filter intramolecular if requested (default) | |
if not intramolecular: | |
vp_chain = residue.chain.index | |
res_nl = [r for r in res_nl if r.chain.index != vp_chain] | |
# 2d. Expand selection to all atoms of selected residues | |
nl = {a for r in res_nl for a in r.atoms} | |
# 2e. Generate reduced trajectory with only these atoms. | |
universe_idx = {a.index for a in nl} | {a.index for a in residue.atoms} | |
universe = t.atom_slice(list(universe_idx), inplace=False) | |
# Create hbond matrix object | |
# Make mapping between residue index and hbond row index | |
mapping = {r.index: idx for idx, r in enumerate(universe.top.residues)} | |
mapping_r = {idx: r for idx, r in enumerate(universe.top.residues)} | |
# Shape: ResxFrame | |
hbond_mtx = np.zeros((len(mapping), t.n_frames), dtype=np.int32) | |
# Atom to Res dictionary | |
atomres_dict = {a.index: a.residue for a in universe.top.atoms} | |
# 3. Get hydrogen bonds between residue and these residues | |
vp_idx = residue.index | |
for f_idx, f in enumerate(universe): | |
hb_list = md.baker_hubbard(f, freq=0.0) | |
for hb in hb_list: | |
d, _, a = hb | |
d_res, a_res = atomres_dict.get(d), atomres_dict.get(a) | |
if d_res.index == vp_idx: | |
other = a_res | |
elif a_res.index == vp_idx: | |
other = d_res | |
else: | |
continue | |
o_idx = mapping[other.index] | |
hbond_mtx[o_idx, f_idx] += 1 | |
# Remove empty rows in the matrix | |
zero_rows = {idx for idx, c in enumerate(hbond_mtx.sum(axis=1)) if c == 0} | |
# Update indexes to create labeling dict (idx to residue obj) | |
mapping_r = {idx: r for idx, r in mapping_r.items() if idx not in zero_rows} | |
sorted_keys = sorted(mapping_r.keys()) | |
mapping_r = {idx: mapping_r[key] for idx, key in enumerate(sorted_keys)} | |
hbond_mtx = np.delete(hbond_mtx, sorted(zero_rows), 0) | |
return (mapping_r, hbond_mtx) | |
def plot_barcode(data, labels): | |
"""Creates a barcode plot from a 2D array | |
data: 2D array (preferrably binary values) | |
labels: dictionary with row idx to residue object | |
""" | |
n_strips = len(labels) | |
fig, axes = plt.subplots(nrows=n_strips, ncols=1) | |
barprops = dict(aspect='auto', cmap='Reds', interpolation='nearest') | |
for idx, ax in zip(labels, axes): | |
data_strip = data[idx, :].reshape(1, -1) | |
ax.imshow(data_strip, **barprops) | |
ax.yaxis.set_major_locator(plt.NullLocator()) | |
ax.xaxis.set_major_locator(plt.NullLocator()) | |
ax.xaxis.set_minor_locator(plt.NullLocator()) | |
# Y labels | |
ax.set_yticklabels([]) | |
r = labels[idx] | |
labels_str = f'{r.chain.index}:{r.code}{r.resSeq}' | |
ax.set_ylabel(labels_str, rotation="horizontal", ha='right', va='center_baseline') | |
# Design | |
ax.grid(False) | |
# Set parameters for last ax x-axis | |
ax.xaxis.set_major_locator(plt.AutoLocator()) | |
ax.set_xlabel('Simulation Frame') | |
fig.align_ylabels() | |
fig.tight_layout() |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment