If you have a header line in your data, you might want to consider to use a hash of arrays. Each array would contain one of your columns and could be adressed by the name given in the header.

use strict;
use warnings;
use Data::Dumper;

my $n = 4;    # number of columns, given as user input
my %hcolumns; # this is the hash of the columns you are looking for

my @header = split /,/,<DATA>; # pls use Text::CSV for any real work
chomp @header;
while(<DATA>){
        chomp;
        my @record = split /,/;    # pls use Text::CSV for any real wo
+rk
        for (0..$n-1) { 
                push @{ $hcolumns{$header[$_]} }, $record[$_];        
+ }
}
print Dumper \%hcolumns; # hash of columns

__DATA__
Name,FirstName,Instrument
Mangelsdorff,Albert,Trombone
Parker,Charlie,Trumpet
Coltrane,John,Tenor
Evans,Bill,Piano
[download]

The number of columns can still be provided by the user. In this example here, I have chosen a number larger that the available columns resulting in an empty column (and some warnings when running the script).