User:Timothee Flutre/Notebook/Postdoc/2012/02/29

From OpenWetWare
Jump to: navigation, search
Owwnotebook icon.png Project name Report.pngMain project page
Resultset previous.pngPrevious entry      Next entryResultset next.png

Use R to extract the best SNP among all in cis of any given gene

This can be reformulated more generically as "extract one row for each level of a column based on the value of a second column":

Let's create a dummy dataset:

x <- data.frame("g"=c("g2","g1","g2","g1"), "s"=c("s1","s2","s3","s4"), p=c(10^-4,10^-3,10^-2,10^-5))

Here is how it looks like:

   g  s     p
1 g2 s1 1e-04
2 g1 s2 1e-03
3 g2 s3 1e-02
4 g1 s4 1e-05

For instance, the "g" column indicates gene names, the "s" column indicates SNP names, and the "p" column indicates P-values of association between genotypes at the SNP and variation in gene expression levels. In such a case, I want to extract the best SNP for each gene, ie. those with the lowest P-value.

First, I sort the "g" column according to the P-values and I keep the row indices after sorting:

v <- x$g[i <- order(x$p)]
[1] g1 g2 g1 g2
[1] 4 1 2 3

Second, I find the first occurrence of each level in column "g" (in this example, the first occurrence corresponds to the lowest P-value for this gene):

min.occ <- !duplicated(v)

Third, I retrieve the row indices corresponding to these first occurrences per level:

row.idx <- setNames(seq_len(nrow(x))[i][min.occ], v[min.occ])
g1 g2
 4  1

Finally, I extract the data I'm interested in:

   g  s     p
4 g1 s4 1e-05
1 g2 s1 1e-04

This could also be done more simply by:

   g  s     p
4 g1 s4 1e-05
1 g2 s1 1e-04

And according to this answer on SO, the whole procedure seems pretty fast!