Extracting pdb structures

[GSINGH006]

Hi, is there a method in Deeptime similar to Pyemma's " pyemma.coordinates.save_traj" which can be used to extract pdb structures representing macrostates after coarse-graining using PCCA+

[clonker]

There is no direct MD dependency in deeptime so in that sense, no, but you can use deeptime to estimate your models and extract the indices, put these into pyemma :)

[GSINGH006]

Hi @clonker I am using the following code for the same

topfile = "/home/lab/Downloads/penta_traj/pentapeptide-impl-solv.pdb" traj_list=["/home/lab/Downloads/penta_traj/pentapeptide-00-500ns-impl-solv.xtc","/home/lab/Downloads/penta_traj/pentapeptide-01-500ns-impl-solv.xtc"] import pyemma import deeptime as dt import numpy as np import matplotlib.pyplot as plt import mdshare plt.matplotlib.rcParams.update({'font.size': 16}) import numpy as np feat1= pyemma.coordinates.featurizer(topfile) tor=feat1.add_backbone_torsions(cossin=True) y= pyemma.coordinates.load(traj_list, feat1) from deeptime.decomposition import TICA tica_estimator = TICA(lagtime=1,dim=2) tica = tica_estimator.fit(y).fetch_model() tics=tica.transform(y) from tqdm.notebook import tqdm from deeptime.clustering import KMeans kmeans_estimator = KMeans(n_clusters=30, progress=tqdm) clustering = kmeans_estimator.fit(np.concatenate(tics)[::1]).fetch_model() fig, ax = plt.subplots() ax.plot(clustering.cluster_centers.T, 'k+') pyemma.plots.plot_free_energy(np.concatenate(tics).T, ax=ax) plt.savefig('energy_dist.png', dpi=300) dtrajs = [] for projected_trajectory in tics: dtrajs.append(clustering.transform(projected_trajectory)) **msm = MaximumLikelihoodMSM(lagtime=50).fitfetch(dtrajs) nstates = 2 cg = msm.pcca(nstates) for i, s in enumerate(cg.sets): print('π{} = {:f}'.format(i + 1, msm.stationary_distribution[s].sum())) pcca_dist1=cg.metastable_distributions indices = dt.markov.sample.compute_index_states(dtrajs) ind=dt.markov.sample.indices_by_distribution(indices,distributions=pcca_dist1,nsample=1) pyemma.coordinates.save_traj(traj_list, ind[0],'/home/lab/samples_pcca1.xtc' , top=topfile) Is this the correct way to extract the indices and use them for getting the representative structures? what does nsamples in this function exactly represent ? Thank you