def featurize_file(job_tuple):
yaml_file, protein, feat, traj_file,stride = job_tuple
yaml_file = load_yaml_file(yaml_file)
if feat is None:
feat = DihedralFeaturizer(types=['phi', 'psi','chi1'])
_check_output_folder_exists(yaml_file, protein)
output_folder = os.path.join(yaml_file["base_dir"],
protein,
yaml_file["feature_dir"])
traj_name = os.path.splitext(os.path.basename(traj_file))[0]
output_fname = os.path.join(output_folder, traj_name+".jl")
feat_descriptor = os.path.join(output_folder, "feature_descriptor.h5")
try:
trj = mdt.load(traj_file)
except :
warnings.warn("Removing %s because of misformed trajectory"%traj_file)
os.remove(traj_file)
return
features = feat.partial_transform(trj)
verbosedump(features, output_fname)
if not os.path.isfile(feat_descriptor) and hasattr(feat, "describe_features"):
dih_df = pd.DataFrame(feat.describe_features(trj[0]))
verbosedump(dih_df, feat_descriptor)
return
def test_DihedralFeaturizer_describe_features_nosincos():
feat = DihedralFeaturizer(sincos=False)
rnd_traj = np.random.randint(len(trajectories))
features = feat.transform([trajectories[rnd_traj]])
df = pd.DataFrame(feat.describe_features(trajectories[rnd_traj]))
for f in range(25):
f_index = np.random.choice(len(df))
atom_inds = df.iloc[f_index].atominds
feature_value = md.compute_dihedrals(trajectories[rnd_traj],
[atom_inds])
if feat.sincos:
func = getattr(np, '%s' % df.iloc[f_index].otherinfo)
feature_value = func(feature_value)
assert (features[0][:, f_index] == feature_value.flatten()).all()
def test_DihedralFeaturizer_describe_features_nosincos():
feat = DihedralFeaturizer(sincos=False)
rnd_traj = np.random.randint(len(trajectories))
features = feat.transform([trajectories[rnd_traj]])
df = pd.DataFrame(feat.describe_features(trajectories[rnd_traj]))
for f in range(25):
f_index = np.random.choice(len(df))
atom_inds = df.iloc[f_index].atominds
feature_value = md.compute_dihedrals(trajectories[rnd_traj],
[atom_inds])
if feat.sincos:
func = getattr(np, '%s' % df.iloc[f_index].otherinfo)
feature_value = func(feature_value)
assert (features[0][:, f_index] == feature_value.flatten()).all()
def Get_dihedral_features_villin():
import os
import shutil
import mdtraj as md
os.chdir('/homes/anuginueni/traj_villin')
if(os.path.isdir('./diheds')):
shutil.rmtree('./diheds')
from msmbuilder.dataset import dataset
t=md.load( "/homes/anuginueni/traj_villin/trajectory-331.xtc",top='/homes/anuginueni/traj_villin/filtered.pdb',stride=5)
xyz = dataset( "/homes/anuginueni/traj_villin/*.xtc",topology='/homes/anuginueni/traj_villin/filtered.pdb',stride=5)
from msmbuilder.featurizer import DihedralFeaturizer #for dihedrals
featurizer = DihedralFeaturizer(types=['phi', 'psi']) #for dihedrals
diheds = xyz.fit_transform_with(featurizer, 'diheds/', fmt='dir-npy') #for dihedrals
des_feat=featurizer.describe_features(t)
res = [ sub['resids'] for sub in des_feat ]
print(str(res))
return diheds
def test_FeatureSelector_describe_features():
rnd_traj = np.random.randint(len(trajectories))
f_ca = ContactFeaturizer(scheme='CA', ignore_nonprotein=True)
f1 = f_ca.transform([trajectories[rnd_traj]])
df1 = pd.DataFrame(f_ca.describe_features(trajectories[rnd_traj]))
f_dih = DihedralFeaturizer()
f2 = f_dih.transform([trajectories[rnd_traj]])
df2 = pd.DataFrame(f_dih.describe_features(trajectories[rnd_traj]))
df_dict = {}
df_dict["ca"] = df1
df_dict["dih"] = df2
f_comb = FeatureSelector([('ca', f_ca), ('dih', f_dih)])
f3 = f_comb.transform([trajectories[rnd_traj]])
df3 = pd.DataFrame(f_comb.describe_features(trajectories[rnd_traj]))
assert len(df3) == len(df1) + len(df2)
df4 = pd.concat([df_dict[i] for i in f_comb.feat_list])
# lets randomly compare 40 features
for i in np.random.choice(range(len(df3)), 40):
for j in df3.columns:
assert eq(df3.iloc[i][j], df4.iloc[i][j])
def test_FeatureSelector_describe_features():
rnd_traj = np.random.randint(len(trajectories))
f_ca = ContactFeaturizer(scheme='CA', ignore_nonprotein=True)
f1 = f_ca.transform([trajectories[rnd_traj]])
df1 = pd.DataFrame(f_ca.describe_features(trajectories[rnd_traj]))
f_dih = DihedralFeaturizer()
f2 = f_dih.transform([trajectories[rnd_traj]])
df2 = pd.DataFrame(f_dih.describe_features(trajectories[rnd_traj]))
df_dict = {}
df_dict["ca"] = df1
df_dict["dih"] = df2
f_comb = FeatureSelector([('ca', f_ca), ('dih', f_dih)])
f3 = f_comb.transform([trajectories[rnd_traj]])
df3 = pd.DataFrame(f_comb.describe_features(trajectories[rnd_traj]))
assert len(df3) == len(df1) + len(df2)
df4 = pd.concat([df_dict[i] for i in f_comb.feat_list])
# lets randomly compare 40 features
for i in np.random.choice(range(len(df3)), 40):
for j in df3.columns:
assert eq(df3.iloc[i][j], df4.iloc[i][j])
def featurize_traj(job_tuple):
#separate out the job tuple into required things
mutant,mutant_dir,project,proj_folder,proj_top_folder,traj_file,stride,save_common,allowed_residue_ind \
= job_tuple
#load top file to setup solute/solvent indices
top_path = os.path.join(proj_top_folder, "%s.pdb"%os.path.basename(traj_file).split("_")[0])
top_trj = mdtraj.load(top_path)
#set up featurizer objects
dihedral_feat = DihedralFeaturizer(types=['phi', 'psi','chi1'])
#load the trajectory
try:
trj = mdtraj.load(traj_file,stride=stride)
except:
print "Cant featurize %s"%traj_file
return
#setup file name
traj_name = os.path.splitext(os.path.basename(traj_file))[0]
print traj_name
dihedral_output_file = os.path.join(mutant_dir,"features/dihedral_features/")+str(project)+\
"_"+traj_name+".h5"
water_output_file = os.path.join(mutant_dir,"features/water_features/")+str(project)+\
"_"+traj_name+".h5"
combined_output_file = os.path.join(mutant_dir,"features/combined_features/")+str(project)+\
"_"+traj_name+".h5"
do_again=True
already_done=False
if os.path.isfile(combined_output_file):
f = verboseload(combined_output_file)
if f.shape[0]!=trj.n_frames:
already_done=True
if not already_done or do_again:
dihedral_features = dihedral_feat.partial_transform(trj)
traj_name = os.path.splitext(os.path.basename(traj_file))[0]
dihedral_output_file = os.path.join(mutant_dir,"features/dihedral_features/")+str(project)+\
"_"+traj_name+".h5"
#now we can dump
verbosedump(dihedral_features,dihedral_output_file)
if save_common:
dih_df = pandas.DataFrame(dihedral_feat.describe_features(top_trj))
dih_f_ind = numpy.array([set(i).issubset(allowed_residue_ind) for i in dih_df["resid"]])
subset_dihedral_features = dihedral_features[:,dih_f_ind]
dihedral_output_file = os.path.join(mutant_dir,"features/common_basis/dihedral_features/")+\
str(project)+"_"+traj_name+".h5"
#now we can dump
verbosedump(subset_dihedral_features,dihedral_output_file)
#save the featurizer information.
verbosedump([dih_df,allowed_residue_ind,dih_f_ind,],\
os.path.join(mutant_dir,"features/common_basis/dihedral_features/")+"saved_dihed_feat.h5")
return
else:
print "skipping featurization for %s since its already done"%traj_name
return
top = md.load("../top.pdb")
trj_list = [md.load(i, top=top) for i in flist]
print("Found %d trajs" % len(trj_list))
f = DihedralFeaturizer(sincos=False)
dump(f, "raw_featurizer.pkl")
feat = f.transform(trj_list)
dump(feat, "raw_features.pkl")
f = load("./featurizer.pkl")
dump(f, "featurizer.pkl")
df1 = pd.DataFrame(f.describe_features(trj_list[0]))
dump(df1, "feature_descriptor.pkl")
feat = f.transform(trj_list)
dump(feat, "features.pkl")
t = tICA(lag_time=100, n_components=2, kinetic_mapping=False)
tica_feat = t.fit_transform(feat)
dump(t, "tica_mdl.pkl")
dump(tica_feat, "tica_features.pkl")
kmeans_mdl = KMeans(50)
ass = kmeans_mdl.fit_predict(tica_feat)
msm_mdl = MarkovStateModel(100)
#msmbuilder imports
from msmbuilder.dataset import dataset
from msmbuilder.featurizer import ContactFeaturizer
from msmbuilder.featurizer import DihedralFeaturizer
from msmbuilder.decomposition import tICA
from msmbuilder.cluster import MiniBatchKMeans
from msmbuilder.msm import ContinuousTimeMSM
from msmbuilder.utils import verbosedump, verboseload
from msmbuilder.cluster import KCenters
from msmbuilder.utils import load, dump
#other imports
import os, glob, shutil
import numpy as np
import mdtraj as md
import pandas as pd
import pickle
#prettier plots
a = np.arange(1119, 1277)
top = md.load("../prot.pdb", atom_indices=a)
# swap this for whatever you have. The code for now supports contacts, dihedral, and angles.
feat = DihedralFeaturizer(types=['chi1', 'chi2'])
# this basically maps every feature to atom indices.
df1 = pd.DataFrame(feat.describe_features(top))
dump(df1, "feature_descriptor.pkl")
featurizer = DihedralFeaturizer(types=['chi1', 'chi2']) #dump(featurizer,"raw_featurizer.pkl") #from msmbuilder.utils import load,dump f = DihedralFeaturizer(types=['chi1', 'chi2'], sincos=False) dump(f, "raw_featurizer.pkl") #featurizer = DihedralFeaturizer(types=['chi1', 'chi2'], resids= 73,74,75,76,77,78,79,80,81,82,83) diheds = featurizer.fit_transform(ds) dump(diheds, "features.pkl") #print(ds[0].shape) print(diheds[0].shape) # this basically maps every feature to atom indices. df1 = pd.DataFrame(featurizer.describe_features(ds)) dump(df1, "feature_descriptor.pkl") #Robust scaling from msmbuilder.preprocessing import RobustScaler scaler = RobustScaler() scaled_diheds = scaler.fit_transform(diheds) print(diheds[0].shape) print(scaled_diheds[0].shape) #Reducing dimension tica_model = tICA(lag_time=1, n_components=10) # fit and transform can be done in seperate steps: tica_model.fit(diheds) tica_trajs = tica_model.transform(diheds)
