pca analysis for specific atoms

[Former user]

Hi There,

Could anybody suggest me a way to do a pca analysis for specific atoms from an ensemble of protein structures?

I have a series of dummy atoms in each protein that correspond to the centre of mass of the ligand. I would like to cluster these dummy atoms on the protein space to see their distribution.

So, after aligning the protein structures, I am wondering how to do a pca/clustering for these dummy atoms alone.

Thanks, Subha

[Xinqiu Yao]

Hi Subha,

Maybe others can give an even better method, but one solution to your problem is to use the read.all() function. Before calling 'read.all()', make sure all your structures are pre-fitted. For example,

# fit based on core/all-CA atoms and generate fitted all-atom structures into 'fitlsq'.
xyz <- fit.xyz(pdbs$xyz[1, ], pdbs, fixed.inds=core$xyz, mobile.inds=core$xyz, full.pdbs=TRUE, ncore=NULL)

# read all atom structure information
pdbs$id <- paste('fitlsq/', pdbs$id, '_flsq.pdb', sep='')
pdbs.aa <- read.all(pdbs)

# ligand atoms are stored in `pdbs.aa$all.hetatm`
# make 'xyz' for dummy; Suppose dummy atoms are named 'DU'
dummy.xyz <- sapply(pdbs.aa$all.hetatm, function(x) {
  inds <- atom.select(x, elety='DU')
  x$xyz[inds$xyz]
})
dummy.xyz <- as.xyz(t(dummy.xyz))
dummy.xyz

# do PCA for dummy atoms
pc <- pca(dummy.xyz)

Unfortunately, I don't have suitable examples to test above codes. Let me know if you have any question/problem of running it.

[skal24]

Hi Yao,

Thanks for the suggestion. Will check that out.

Regards, Subha