Tutorials adapted from MDtraj¶
try pytraj online:
Contents
Clustering with scipy¶
Original tutorial is from MDtraj
Note
this example only shows how to populate pytraj to python‘s ecosystem. Users should check Cluster Analysis
In [1]:
from __future__ import print_function
import warnings
warnings.filterwarnings('ignore', category=DeprecationWarning)
%matplotlib inline
import pytraj as pt
import scipy
import scipy.cluster.hierarchy
In [2]:
# load data
traj = pt.iterload('tz2.nc', 'tz2.parm7')
traj
Out[2]:
In [3]:
# calculate pairwise rmsd with `autoimage=True`
# generally we only need to cluster a subset of atoms.
# cluster for 'CA' atoms
distances = pt.pairwise_rmsd(traj(autoimage=True), mask='@CA')
# use `scipy` to perform clustering
linkage = scipy.cluster.hierarchy.ward(distances)
scipy.cluster.hierarchy.dendrogram(linkage, no_labels=True, count_sort='descendent')
None
In [4]:
# cluster for all atoms but H
distances = pt.pairwise_rmsd(traj(autoimage=True), mask='!@H=')
linkage = scipy.cluster.hierarchy.ward(distances)
scipy.cluster.hierarchy.dendrogram(linkage, no_labels=True, count_sort='descendent')
None
(clustering_mdtraj_adapted.ipynb; clustering_mdtraj_adapted_evaluated.ipynb; clustering_mdtraj_adapted.py)
Principal component analysis with sklearn¶
Original tutorial is from here
Note
cpptraj (then pytraj) has PCA analysis. Doc is coming soon.
PCA analysis with sklearn (adapted from mdtraj and cpptraj tutorial)¶
In [1]:
import warnings
warnings.filterwarnings('ignore', category=DeprecationWarning)
%matplotlib inline
from __future__ import print_function
%config InlineBackend.figure_format = 'retina' # high resolution
import matplotlib
matplotlib.rcParams['savefig.dpi'] = 2 * matplotlib.rcParams['savefig.dpi'] # larger image
import pytraj as pt
import matplotlib.pyplot as plt
from sklearn.decomposition import PCA
from pytraj.plot import show_config
In [2]:
# we use `load` method to load all data to memory. This is good for small data size.
# use `pytraj.iterload` for out-of-core traj.
traj = pt.load('tz2.nc', 'tz2.parm7')
traj
Out[2]:
In [3]:
pca = PCA(n_components=2)
pca
Out[3]:
In [4]:
# superpose to 1st frame
pt.superpose(traj, ref=0, mask='!@H=')
Out[4]:
In [5]:
# create average structure
avg = pt.mean_structure(traj)
avg
Out[5]:
In [6]:
# superpose all structures to average frame
pt.superpose(traj, ref=avg, mask='!@H=')
Out[6]:
In [7]:
# perform PCA calculation and get transformed coords
# we need to reshape 3D traj.xyz array to 2D to make sklearn happy
# make a new traj by stripping all H atoms
traj_new = traj['!@H=']
xyz_2d = traj_new.xyz.reshape(traj_new.n_frames, traj_new.n_atoms * 3)
print(xyz_2d.shape) # (n_frames, n_dimensions)
In [8]:
reduced_cartesian = pca.fit_transform(xyz_2d)
print(reduced_cartesian.shape) # (n_frames, n_dimensions)
In [9]:
# ignore warning
import warnings
warnings.filterwarnings('ignore')
plt.figure()
plt.scatter(reduced_cartesian[:, 0], reduced_cartesian[:,1], marker='o', c=range(traj_new.n_frames), alpha=0.5)
plt.xlabel('PC1')
plt.ylabel('PC2')
cbar = plt.colorbar()
cbar.set_label('frame #')
Compare to cpptraj data¶
note: stop here if you do not care (a bit compilicated code)
In [10]:
# cpptraj
# copy from Amber15 manual (page 619)
command = '''
# Step one. Generate average structure.
# RMS-Fit to first frame to remove global translation/rotation.
parm tz2.parm7
trajin tz2.nc
rms first !@H=
average crdset AVG
run
# Step two. RMS-Fit to average structure. Calculate covariance matrix.
# Save the fit coordinates.
rms ref AVG !@H=
matrix covar name MyMatrix !@H=
createcrd CRD1
run
# Step three. Diagonalize matrix.
runanalysis diagmatrix MyMatrix vecs 2 name MyEvecs
# Step four. Project saved fit coordinates along eigenvectors 1 and 2
crdaction CRD1 projection evecs MyEvecs !@H= out project.dat beg 1 end 2
'''
In [11]:
state = pt.datafiles.load_cpptraj_state
In [12]:
state = pt.datafiles.load_cpptraj_state(command)
# tell 'run' to perform all calculation
state.run()
Out[12]:
In [13]:
# get data
state.data
Out[13]:
In [14]:
print([dset.key for dset in state.data])
print(state.data['MyMatrix'].values.shape)
In [15]:
# reduced_cartesian corresponds to dataset with names of 'Mode1', 'Mode2'
mode_0, mode_1 = -state.data['Mode1'].values, -state.data['Mode2'].values
# mode_0, mode_1 = state.data['Mode1'].values, state.data['Mode2'].values
# plot: cpptraj
fig = plt.figure()
ax_0 = fig.add_subplot(211)
ax_0.scatter(mode_0, mode_1, marker='o', c=range(traj.n_frames), alpha=0.5)
ax_0.set_xlabel('PC1')
ax_0.set_ylabel('PC2')
ax_0.set_xlim([-60, 80])
ax_0.set_ylim([-40, 40])
ax_0.set_title('cpptraj-flip')
ax_0.set_yticks([-40, -20, 0, 20, 40])
# plot: sklearn
fig = plt.figure()
ax_1 = fig.add_subplot(212)
ax_1.scatter(reduced_cartesian[:, 0], reduced_cartesian[:,1], marker='o', c=range(traj.n_frames), alpha=0.5)
ax_1.set_xlabel('PC1')
ax_1.set_ylabel('PC2')
ax_1.set_yticks([-40, -20, 0, 20, 40])
ax_1.set_title('sklearn')
Out[15]:
In [16]:
print('sklearn \n')
print(reduced_cartesian[:, 0], reduced_cartesian[:, 1])
print('\ncpptraj\n') # flip sign, why?
print(mode_0, mode_1)
(pca_mdtraj_adapted.ipynb; pca_mdtraj_adapted_evaluated.ipynb; pca_mdtraj_adapted.py)