0

I have a matrix where I have counts for genes in different samples

Col1: GeneName
Col2: Length
Col3;Col4;Col5; Counts for genes in sampleA/sampleB/sampleC
Col6;Col7;Col8; Total counts in sampleA/sampleB/sampleC

This is an example matrix.

A1BG    1758    53  4373    207 46005749    43849471    31554941 
A1BG-AS1    2126    5   88  12  46005749    43849471    31554941
A1CF    9695    8882    3522    437 46005749    43849471    31554941 
A2M 5399    15963   12325   7227    46005749    43849471    31554941 
A2M-AS1 6660    50  33  36  46005749    43849471    31554941 

I want to divide counts_sampleA / (total_counts_sampleA*Length), and so on for other samples

cat inFile | awk 'BEGIN {OFS="\t"} { print $1,$2,$3/($6*$2),$4/($7*$2),$5/($8*$2) }'

This is the expected result

A1BG    1758    6.55307e-10 5.67278e-08 3.73151e-09  
A1BG-AS1    2126    5.11204e-11 9.43963e-10 1.78875e-10   
A1CF    9695    1.99136e-08 8.28471e-09 1.42845e-09   
A2M 5399    6.42672e-08 5.20606e-08 4.24207e-08   
A2M-AS1 6660    1.63186e-10 1.12999e-10 1.71301e-10  

Works fine, but isn't good when the matrix is large. How would I write if there are 100 samples, where column3-colum102 would have geneCountinEachSample and Coulmn103-column202 would have totalCountinEachSample.

I want to use it with a for loop, so when there are more samples, it works with an arbitrary number of columns ?

cat inFile | awk 'BEGIN {OFS="\t"} { row=NF; samples=3; size=$samples+2; for ( i=3; i<=$size; i++); END print $i/$[$i+$samples] }'

Any suggestion on how to make this work. Thanks !

6
  • Is the sample always 3? As when say it is 4 or more, then you run out of columns to divide with. Also, your comments "(col3/col6)*col2" DONOT match the awk expression "$3/($6*$2)". That is why you need to supply input and expected output for various values of sample.
    – user218374
    Apr 29, 2017 at 0:22
  • You shouldn't write $samples, use samples instead. Using a dollar sign gives you the samplesth column's value. Apr 29, 2017 at 7:22
  • In summary, please define your problem clearer. Apr 29, 2017 at 7:23
  • Sorry, i have tried to make the question clear.
    – CN_229133
    Apr 29, 2017 at 8:01
  • show the expected result Apr 29, 2017 at 8:07

2 Answers 2

1

Well, you almost got the answer there:

awk '
     {cols=((NF/2) + 1)
      for (i=1; i <= cols; i++) {
          if (i >= 3) {
              count_index= i + cols - 2
              printf("%s\t", 1.0 * $i / ($count_index * $2))
          } else {
              printf("%s\t", $i) 
          }
      }
      printf("\n")
     }' inFile

Note that using cat file | awk ... is suboptimal, awk handles files as arguments directly; even so, doing awk ... < infile would be better than a useless use of cat.

0
perl -F'\s+' -lane '$,="\t"; # OFS made a TAB
   my($gN, $gL) = splice @F, 0, 2; # store gene name & length
   print $gN, $gL, map { sprintf "%.5e", $F[$_] / ( $F[$_+@F/2] * $gL ) } 0 .. @F/2-1;
' gene_samples.file

  • FS is set to one or more whitespace.
  • ORS = RS = \n
  • @F holds fields for a given record.
  • splice breaks away 2 elements starting from offset 0 and reduces the array size as well.
  • From OP spec, from what remains in @F is even numbered elements. The first half is counts_for_each_sample & the latter half is total_count_for_each_sample.

Results

A1BG      1758  6.55307e-10  5.67278e-08  3.73151e-09
A1BG-AS1  2126  5.11204e-11  9.43963e-10  1.78875e-10
A1CF      9695  1.99136e-08  8.28471e-09  1.42845e-09
A2M       5399  6.42672e-08  5.20606e-08  4.24207e-08
A2M-AS1   6660  1.63186e-10  1.12999e-10  1.71301e-10

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .