You are here

Cluster methodology

1 post / 0 new
Cluster methodology


Questions from a beginner :

Regarding selecting the best clusters, i generated 24000 decoys and the i clustered them with the command:
/rosetta_scripts/cluster_tools/C/cluster_info_silent.out all.out - cluster/a 5,10,15 3,4

Then i made the dendogram that is here :

i'm about 1-2 ansgtrom in the rmsd, i think, but the populations of the clusters are low, how can i populate the clusters more? changing from 15 to like 20 or 30, in the cluster_info_silent.out command? Then choosing the lowest energy cluster and the more populated one?

Then after that if i choose the best cluster, its only a backbone trace right? Whats the best strategy, Rosettarelax, RosettaDesign...? I guess relax does the full-atom refinement or is there another step after this? The cluster that i choose is my starting structure, that i have to put on the paths.txt file? If so then maybe i have to do Relax with repacking on flexible or not flexible backbone?

Last question, Is this the best algorithm for clustering or the in rosetta3 its better, or are there any other options. The scripts from J.Daily Lab?

Any input on this will be appreciated, and a discussion could be started. I also have NMR backbone data available, so i'm trying RosettaNMR also.
I also made plots of the data, what should i plot: score vs rmsd or score vs Rg ( radius of gyration), in the last case should i observe a funnel like in the score vs rmsd case? Rosetta doesn't read my native structure, its an nmr ensemble, i dont know how to change it maybe put just one model? It didn't work, my mistake possibly.

looking forward for discussing this,



Wed, 2009-04-29 11:53