How to extract structure from near the center of the clusters?

[buruiyu]

Hello! I am a new learner. I am using pyemma on linux system. My pyemma version is 2.5.11. After I constructed the model, I found that the representative structures saved inside each macrostate were different, so I first used print(msm.metastable_distributions[1]) to get the distribution of a macrostate over each microstate, and selected the microstate with the highest probability; then used msm.sample_by_state( 10 , [1]) to extract 10 structure from inside a microstate, and then use pyemma.coordinates.save_trajs to save it. I extracted 10 structures from inside each microstate using sample_by_state, what is the criteria for extracting structures with this line of code? Is each structure extracted sorted by some weight? After I extracted it this way, the gap between each structure inside the microstates is still large, so I want to extract some structures near the cluster centroid of the microstates, how should I implement it?

Thank you!

[clonker]

Hey, sample by state does not use any weighting, the structures are drawn uniformly from the pool of frames that are assigned to the particular state. If they deviate a lot, it could mean that either your projection isn't good or the clustering doesn't use enough centers (# cluster centers == # states).

Cheers!