overlap_analysis

[karolk]

Hi,

I am trying to allign two conformational states, using coordinates of a fixed domain, and then calculate difference vector between these two for the whole structures, and then compare this diff.vector to my normal modes.

I tried the following:

1

aln <- struct.aln(closed, open, fixed.inds = closed.inds,mobile.inds = open.inds, max.cycles=0)
closed$xyz <- aln$xyz
xyz <- rbind(open$xyz[aln$a.inds$xyz], closed$xyz[aln$a.inds$xyz])
diff <- difference.vector(xyz)
oa <- overlap(modes, diff)

Then, I am getting error: Error in overlap(modes, diff) : overlap: unequal vector lengths

2

• I aligned my structures, using a different program

• I load these in R

• I calculated diff.vector:

- xyz <- rbind(open$xyz,closed$xyz)
- diff <- difference.vector(xyz)
- oa <- overlap(modes,diff)

Again, the same error: Error in overlap(modes, diff) : overlap: unequal vector lengths

I know what does it mean, but I don't know why.

thanks

[Lars Skjærven]

The dimensions of xyz are not the same as your modes object. Your problem is (probably)open$xyz[aln$a.inds$xyz]. Here a.inds$xyz corresponds to atoms used for superposition. aln$xyz contains the superimposed coordinates. read the docu for this function.

Note that for comparing a structure difference vector, you should superimpose on all atoms, and not a subset as the struct.aln() function does. set max.cycles to 0, or use pdbfit or fit.xyz.

[karolk]

Hi Lars,

Yes. You are right. Thats way instead of going with option "1", I decided to try option "2" where I compare structures, the whole structures, which are already aligned. The idea here was simple: load two structures and calculate difference vector without aligning them. I will play with different options and I will let you know.

[Lars Skjærven]

Sounds good. So whats the dimensions of modes$U and xyz (dim(modes$U) and dim(xyz))?

[karolk]

I will have to check:). I spent last 2 hours doing ensemble NMA. Amazing tool:). I will let you know tomorrow.

[karolk]

Hi,

Thank you for the suggestions, regarding dim (modes and xyz). Actually, my xyz was empty. That's way it did not work. Nevertheless, I still was not able to get my difference vector, using coordinates, which were aligned in a different program and loaded in R.

So, option 1: align in R. Here is my code, which uses fit.xyz. I am able to align and computed difference vector for whole proteins, but not for the proteins, which are aligned, using only part of a structure. Actually, fit.xyz fails during alignment.

open <- read.pdb("open.pdb")
closed <- read.pdb("closed.pdb")

sele_open_CA <- atom.select(open, elety='CA')
sele_closed_CA <- atom.select(closed, elety='CA')

# select CAs of  domains, which will be used for alignment
fitDomain_open <- atom.select(open, chain=c("A","J","K"...), elety = "CA")
fitDomain_closed <- atom.select(closed, chain=c("A","J","K"...), elety = "CA")

# bind 
xyz <- rbind(open$xyz[sele_open_CA$xyz], closed$xyz[sele_closed_CA$xyz])

## Superimpose, using coordinates of fitDomain
xyz[2,] <- fit.xyz(xyz[1,], xyz[2,], 1:ncol(xyz),fitDomain_open, fitDomain_closed)

And here comes the error: "Error in fit.xyz(xyz[1, ], xyz[2, ], 1:ncol(xyz), fitDomain_open, fitDomain_closed) : length of 'fixed.inds' != length of 'mobile.inds'" It can't be since my fitDomain_open and fitDomain_closed have the same number of elements. I checked:

fitDomain_open: Call: atom.select.pdb(pdb = open, elety = "CA", chain = c("A", "J", "K", ...))

Atom Indices#: 2224 ($atom) XYZ Indices#: 6672 ($xyz)

fitDomain_closed

Call: atom.select.pdb(pdb = closed, elety = "CA", chain = c("A", "J", "K", ...))

Atom Indices#: 2224 ($atom) XYZ Indices#: 6672 ($xyz)

Thanks

[karolk]

Hi,

I refresh this. For reminder, I am trying to align two different conformations of my protein, using fit.xyz and then calculate diff.vector, and then perform overlap analysis. I want to align, using part of a protein only, and this is where the problems start.For the last few days I have tried a number of different options. None of these worked for me.

Any suggestions will be appreciated.

Example of another code. In this case the error was: Error in fit.xyz(xyz[1, ], xyz[2, ], 1:ncol(xyz), fixed.inds = caTM.inds.open$xyz, : NA elements selected for fitting (check indices)

open <- read.pdb("X.pdb")
clos <- read.pdb("Y.pdb")

ca.inds.open <- atom.select(open, elety='CA')
ca.inds.clos <- atom.select(clos, elety='CA')

caTM.inds.open <- atom.select(open, chain=c("X","Y","Z"), elety='CA')
caTM.inds.clos <- atom.select(clos, chain=c("X","Y","Z"), elety='CA')

xyz <- rbind(ca.inds.open$xyz,ca.inds.clos$xyz)

xyz[2,] <- fit.xyz(xyz[1,], xyz[2,], 1:ncol(xyz), fixed.inds=caTM.inds.open$xyz, mobile.inds = caTM.inds.clos$xyz)

[Lars Skjærven]

Hello, What's your third input argument to fit.xyz() doing? Also, fitDomain_open and fitDomain_closed are associated with the all-atom open and closed pdb objects, not the CA xyz coordinates which are provided to fit.xyz().

try something in this direction

open <- read.pdb("open.pdb")
closed <- read.pdb("closed.pdb")

sele_open_CA <- atom.select(open, elety='CA')
sele_closed_CA <- atom.select(closed, elety='CA')

open_CA = trim(open, sele_open_CA)
closed_CA = trim(open, sele_closed_CA)

length(open_CA$xyz)
length(closed_CA$xyz)

# select CAs of  domains, which will be used for alignment
fitDomain_open <- atom.select(open_CA, chain=c("A","J","K"))
fitDomain_closed <- atom.select(closed_CA, chain=c("A","J","K"))

# bind 
xyz <- rbind(open_CA$xyz, closed_CA$xyz)

## Superimpose, using coordinates of fitDomain
xyz[2,] <- fit.xyz(xyz[1,], xyz[2,], fitDomain_open$xyz, fitDomain_closed$xyz)

[karolk]

Hi,

Solved. I was aligning, using selections and not the pdb objects=>(trim())

thanks