pattern match and capturing in unix

Question

I have a file like this:

A1-1915435_1_MIMAT0000069_MirBase_miR-16_microRNA_1_22-A1-1915435_1_ENSG00000206737_ENST00000384010_U1_snRNA_73_115
A1-2362643_1_MIMAT0000732_MirBase_miR-378_microRNA_1_21-A1-2362643_1_ENSG00000222328_ENST00000410396_U2_snRNA_168_211

I want an output in which I get micro RNA part and ENSG part tab separated. Some thing like this:

miR-16_microRNA ENSG00000206737
miR-378_microRNA    ENSG00000222328

I can write a perl script for it but I guess there must be command line solution for this using awk, sed perl etc.

Answer 1

perl -pe 's!.*(miR-\d+_microRNA).*(ENSG\d+).*!$1\t$2!'

explain a bit?

Sure, sorry:

• for each line (perl -p)
• substitute regex by $1 tab $2 where
  • $1 and $2 are numbered backreferences to capturing groups (...). Nearly all modern regular expression engines support this.
• about the regular expression:
  • . = any char except \n
  • .* = a sequence of chars (ignored in the substitution)
  • \d = digit (the same as [0-9])
  • \d+ = one or more \d

Please see any reg expression tutorial for a more detailed explanation.

Answer 2

Using sed:

sed 's/^.*\(miR-[0-9]*_microRNA\).*\(ENSG[0-9]*\).*$/\1\t\2/' <infile

Answer 3

Using awk assuming the format is reliable :

$ awk -F'_' '{print $5"_"$6, $11}' file
miR-16_microRNA ENSG00000206737
miR-378_microRNA ENSG00000222328

Answer 4

sed -n 'y/_E/\t\n/;s/\n/E/;s/[^m]*//;s/\(A.\).*E/\1E/;P'

That works for me given your example...

miR-16  microRNA        ENSG00000206737
miR-378 microRNA        ENSG00000222328

If you want the first \tab converted back to a _ then you can add s/\t/_/ before the P - but I think it looks nicer like this.

Another way could look like...

sed 's/RNA[^E]*_/RNA\t/;s/[^m]*//;s/_E.*//'

...or...

sed 's/[^mE]*\([mE][^mE_]*\)/\1/;s//\t\1/3;s/_E.*//'

...which both print what you show in your desired output like...

miR-16_microRNA ENSG00000206737
miR-378_microRNA        ENSG00000222328

. In any case you might have to use a literal character in place of the \t escape depending on your sed implementation.