how to plot the contribution of each residue in PCA histogram.

[Former user]

I was easy to plot the PCA histogram by selecting CA atoms in the pdb file, as described in the bio3d manual. Is there any way to see the contribution of whole residue in the PCA histogram plot?

[Xinqiu Yao]

You can't get contribution of whole residue if your PCA was performed for CA atoms only.

[santosh chaudhary]

HI, thanks for the reply. Yes, but i have performed PCA analysis by taking all atoms of the residues except hydrogen.

[Xinqiu Yao]

You might try the sum of square of loadings within one residue. For example,

tapply(pc$au, pdb$atom$resno, function(x) sqrt(sum(x^2)))

where 'pc' is the pca result and pdb should exactly match the data used for pca.

Let me know if you still have questions.

[Barry Grant]

This can be done for you with the function plot.pca.loadings() it can take a vector of residue numbers as input and will then group your atom contributions for each atom with a similar residue number. Read the docs for this function and give it a try.

[Amir Zeb]

Hello Santosh,

Will you please like to let me know how may I superimpose trajectory taking all the atoms? I know only C_alpha atoms superimposition by command like ca.inds <- atom.select(pdb, elety="CA")

Thanks in advance!

Amir

[Xinqiu Yao]

If use all atoms, you don't need 'atom.select'. Simply call 'fit.xyz()', e.g. xyz.fitted <- fit.xyz(xyz[1, ], xyz).

[Amir Zeb]

Thanks Xin-Qiu,

I tried your instructions but could not produce the pc1 for the whole protein atoms. Once I put the command you suggested, i'm getting some warning like:

xyz.fitted <- fit.xyz(xyz[1, ], xyz)

Warning message: In fit.xyz(xyz[1, ], xyz) : No fitting indices provided, using the 381 non NA positions

Please let me know where am I wrong.

Thanks

[Xinqiu Yao]

The warning is normal - you can safely ignore it. What do you mean 'could not produce the pc1 for the whole protein atoms'? Did pca(xyz.fitted) work?

[Amir Zeb]

Dear Xin-Qiu,

I am following this tutorial "http://thegrantlab.org/bio3d/tutorials/trajectory-analysis" Once I set path to my working directory, I am issuing the following commands.

Library(Bio3d)

dcd <- 'xxx file' pdb <- 'xxx file'

dcd <- read.dcd ("xxx file") pdb <- read.pdb ("xxx file")

xyz.fitted <- fit.xyz(xyz[1, ], xyz) for this command, I am getting the following error

Error in matrix(colSums(rbind(gap.pos1, gap.pos2), na.rm = TRUE), ncol = ncol(x)) : non-numeric matrix extent

Then I tried this command, xyz <- fit.xyz(fixed=pdb$xyz, mobile=dcd), so I got the following warning.

Warning message: In fit.xyz(fixed = pdb$xyz, mobile = dcd) : No fitting indices provided, using the 381 non NA positions

But again if I use the command which you have kindly provided me, I am getting same warning as mentioned in previous comment. If I ignore this warning and use this command p1 <- mktrj.pca(pc, pc=1, b=pc$au[,1], file="pc1.pdb") to produce the pc1, I could not, which is clearly explained on the Bio3d Web link.

Please let me know if there is some tutorial to do PCA of whole atoms of proteins on their trajectory.

Thanks

[santosh chaudhary]

Dear Amir,

Is the number of atoms in pdb and dcd file are same?

regards santosh

[Xinqiu Yao]

Make sure you have executed pc <- pca.xyz(xyz.fitted), which actually did PCA.

[Amir Zeb]

Thanks Santosh,

Yeah, I have converted the same pdb file to dcd file in VMD, and then performed the PCA analysis in RStudio. Actually, I could get PCA but only for CA (Carbon alpha) atoms.

I want to get PCA for the whole protein, means that side chain atoms should be included.

Will you have please any idea, How to do this?

Thanks

Amir

[Amir Zeb]

Thanks Xin-Qiu,

Let me try.

Amir