CODEWITHSUNDEEP
algorithmrdistanceapply
user1357015
user1357015

Reputation: 11676

Making a distance matrix from position information using the R language

Suppose I have a matrix of the following form:

Residue Can.Count SideChain XCoord  YCoord ZCoord
1       MET         1         A 62.935  97.579 30.223
2       THR         2         A 63.155  95.525 27.079
3       GLU         3         A 65.289  96.895 24.308
4       TYR         4         A 64.899  96.220 20.615
8       LYS         8         A 67.593  96.715 18.023
9       LEU         9         A 65.898  97.863 14.816
10       VAL        10         A 67.664  98.557 11.533

Notice that the numbers 5-6-7 are skipped. What I want to do is make a "distance matrix" between each residue to each other residue. In this case, I want to make a 7x7 matrix with with element (1,3) being the distance between those positions.

Now, I realize that I don't need to fill in the bottom half, everything above the diagonal is sufficient. I also see how I could do this using 2 for loops as follows:

 for(i in 1:7) {
  for(j in i:7){
    mymatrix[i,j] <- calcdistance(xyz1,xyz2) #I have the distance function already coded.

 }
}

I realize that it will always be O(n^2) but I am wondering if I can leverage the power of R to make this matrix using an apply statement (or something even more clever)? I have tried doing so but have been unsuccessful somehow. Thank you for your hep!

Upvotes: 1

Views: 371

Answers (1)

mnel
mnel

Reputation: 115382

What you are looking for is the dist function. See ?dist for details.

I'm not clear what you mean by expecting a 7 by 7 matrix, and then for the element [1,3] to refer to the distance between those (after noting there is no 5,6,7). I am taking it to mean you wish to refer to the Can.Count. You can do this by naming the rows and columns and referring to these names.

Assuming your data is a data.frame called residues, the following will work

  • Note this computes the 2-D distance using x-y coordinates, c('XCoord','YCoord'). You could easily make this 3-D by using c('XCoord','YCoord', 'ZCoord').
dist_matrix <-  as.matrix(dist(residues[, c('XCoord','YCoord')], diag = T))
# this gives a 7 by 7 matrix
dist_matrix
##         1        2         3         4        5        6        7
## 1 0.000000 2.065748 2.4513613 2.3883419 4.737453 2.976579 4.829071
## 2 2.065748 0.000000 2.5359132 1.8773814 4.594774 3.604205 5.433609
## 3 2.451361 2.535913 0.0000000 0.7795672 2.311021 1.143637 2.898770
## 4 2.388342 1.877381 0.7795672 0.0000000 2.739099 1.922875 3.620331
## 5 4.737453 4.594774 2.3110206 2.7390986 0.000000 2.047176 1.843368
## 6 2.976579 3.604205 1.1436367 1.9228755 2.047176 0.000000 1.897470
## 7 4.829071 5.433609 2.8987703 3.6203306 1.843368 1.897470 0.000000

# set the dimension names to the Can.Count so we can refer to them
dimnames(dist_matrix) <- list(residues[['Can.Count']],residues[['Can.Count']] )

# now you can refer to the distance between Can.Count 1 and Can.Count 8
dist_matrix['1','8']

## [1] 4.737453

# note that you need to refer to the dimension names as characters, 
# as this is  7 by 7 matrix, so the following will give 
# an (obvious) error message
dist_matrix[1,8]

## Error: subscript out of bounds

Upvotes: 4

Related Questions