assigning the values in matrix in bash

Question

[smrutimayipanda]

I have a matrix consisting of gene name column and log fold change column. I want to change the log fold change values in -1,1 and 0. if log fold change > 0, it should be 1; if logFC < 0, it should be -1 and if there is no values then it should be NA. Please tell me how to assign these values -1,1,0,NA in these matrix? I am comfortable in bash so please tell me in bash only. Thanks in advance.

[Joe]

While this is doable in bash, arithmetic is really not its strong suit, so I would really advise you to get comfortable with a more versatile language such as R or python.

Please also provide some example input data for people to test with at the very least.

[smrutimayipanda]

A3GALT2 9.80e-02  0.295935  3.58e-02
A4GALT 5.58e-02  0.2759222  4.21e-01
A4GNT -5.50e-03  -1.1805802  2.09e-01
AAAS -4.29e-01  -0.122598  2.22e-01
AACS -1.82e-02  -0.0618869  8.14e-02
AADAC 3.22e-02  0.6967785  -4.37e-01
AADACL2 2.97e-02  -1.8886345  -9.67e-03
AADACL3 -1.26e-01  2.3524335  -3.17e-02

This is the test file.

[smrutimayipanda]

Please provide me the bash script for scientific notation Joe. I understood your logic.

[Joe]

I've given you a functional skeleton, it shouldn't be hard for you to adapt it to scientific notation.

[smrutimayipanda]

But i didnt get the thing you said that the E would need to be substituted for a *10^ string. I am saying about this. What to add in the code?

[Joe]

You should experiment for yourself and try things out. The are answers on StackOverflow for this.

But as I said, you'd be better off in a different language.

Answer 1

[Joe]

Question needs more information (such as the format of the input data etc.) as per my comment, but here's the basics of something functional:

#!/bin/bash

while IFS=',' read -r -a array ; do
  if [ -z "${array[1]}" ] ; then
    fc="NA"
  elif (( "${array[1]}" > 0 )); then
    fc=1
  elif (( "${array[1]}" < 0 )); then
    fc=0
  fi
  echo "${array[0]},$fc"
done < $1

Assuming an input file like:

$ cat test.csv
Gene1,1
Gene2,10
Gene3,
Gene4,-1
Gene5,-30

bash scriptname.sh test.csv will yield:

Gene1,1
Gene2,1
Gene3,NA
Gene4,0
Gene5,0

NOTE:

bash cannot do floating point arithmetic (unless you subprocess to bc or something), so if you have floating point data, this is why I would strongly urge the use of Python/R or something as per my comment also. Additionally this is why it's important to provide example input data.

Floating point calculations would require something like the following:

#!/bin/bash

while IFS=',' read -r -a array ; do
  if [ -z "${array[1]}" ] ; then
    fc="NA"
  elif (( $(echo "${array[1]} > 0" | bc)  )); then
    fc=1
  elif (( $(echo "${array[1]} < 0" | bc) )); then
    fc=0
  fi
  echo "${array[0]},$fc"
done < $1

Which for the input:

Gene1,1
Gene2,10
Gene3,
Gene4,-1
Gene5,-30
Gene6,1.234
Gene7,0.999999999999999999999999999999999
Gene8,-0.0000000000000000000000000000000000000000000000000000000000000000000000000000000000000001

will give:

Gene1,1
Gene2,1
Gene3,NA
Gene4,0
Gene5,0
Gene6,1
Gene7,1
Gene8,0

As a final comment, this will still not work for standard scientific notation for exponents (1E+10 etc.). In order for that to work with bc, the E would need to be substituted for a *10^ string.

Answer 2

[Pierre Lindenbaum]

$ echo -e "gene1 100.2\ngene2 -100.4\ngene3" | awk '{OFS="\t";V=($2==""?"NA":($2>=0?1:-1)); print $1,V;}' 
gene1   1
gene2   -1
gene3   NA