Filter a similarity matrix

Question

[ilmagodellepcr]

Dear all,

I obtained a similarity matrix from a multialignment of about 200 protein sequences. I would like to filter this matrix by only 100% identity because I'd like to identify and discard identical accession from this list of proteins.

Any advice?

Thanks in advance

[Devon Ryan]

Just use R, python, perl or any other language that you're familiar with. Note that this may be simpler if the matrix is symmetric, though handling a triangular matrix isn't terribly difficult either.

Answer 1

[newlife.well.2014]

Something like this?

library(data.table)

df = data.frame(p1 = c(1, 0.5, 1), p2= c(0.5, 1, 0.6), p3 = c(1, 0.6, 1))
df1 <- stack(df)
setDT(df1)

df1[, pp := c("p1", "p2", "p3")]

setkey(df1, values)

and the result:

> df1
   values ind pp
1:    0.5  p1 p2
2:    0.5  p2 p1
3:    0.6  p2 p3
4:    0.6  p3 p2
5:    1.0  p1 p1
6:    1.0  p1 p3
7:    1.0  p2 p2
8:    1.0  p3 p1
9:    1.0  p3 p3