How to avoid for loop in this dataset?

Question

For loops in R are extremely slow but I know no alternative way of how to achieve the following.

As shown in this screenshot:

What I want the output format to look like:

> gene_id tss_id x y

in which, x = isosub$q1_FPKM / iso.agg$q1_FPKM // (correspond gene_id)

y = isosub$q2_FPKM / iso.agg$q2_FPKM  

Here is my code with the for loop:

length = length(isosub$gene_id)
tmp = data.frame(isosub$gene_id, isosub$q1_FPKM, isosub$q2_FPKM)
j = 1
denominator_q1 = iso.agg$q1_FPKM[j]
denominator_q2 = iso.agg$q2_FPKM[j]

gene_id = isosub$gene_id
tmpq1 = tmp$isosub.q1_FPKM
tmpq2 = tmp$isosub.q2_FPKM
isoq1 = iso.agg$q1_FPKM
isoq2 = iso.agg$q2_FPKM
o2_q1 = rep(0, length)
o2_q2 = rep(0, length)

i = 0

for (i in 1:length){
     if (gene_id[i+1] == gene_id[i]){
          o2_q1[i] = tmpq1[i] / denominator_q1
          o2_q2[i] = tmpq2[i] / denominator_q2
     }else{
          o2_q1[i] = tmpq1[i] / denominator_q1
          o2_q2[i] = tmpq2[i] / denominator_q2
          j = j + 1
          denominator_q1 = isoq1[j]
          denominator_q2 = isoq2[j]
     }
}

when length = 1000, system.time shows that:

>    user  system elapsed 
>   55.74    0.00   56.45

And my actual length is even larger: 13751.

Answer 1

Do you want to do a merge?

outdf <- merge(isosub[c("gene_id", "tss_id", "q1_FPKM", "q2_FPKM")],
               iso.agg[c("gene_id", "q1_FPKM", "q2_FPKM")],
               by="gene_id",
               suffix=c(".1", ".2"))
outdf$x <- outdf$q1_FPKM.1 / outdf$q1_FPKM.2
outdf$y <- outdf$q2_FPKM.1 / outdf$q2_FPKM.2

Answer 2

If you ended up here looking for ways to avoid or speed up loops, check out this answer: Speed up the loop operation in R

It helped me with a similar problem I was having, and shows ways to keep necessary loops but increase performance dramatically.