Maybe you want something like:

use strict;
use warnings;
use Text::CSV;
use Data::Dumper;

my $csvFile = <<CSV;
A,B,C,D,E,F,G
a1,b,c,"D,1",E1,f,g
a1,b,c,"D,2",E1,f,g
a1,b,c,"D,1",E2,f,g
a2,b,c,"D,1",E3,f,g
a2,b,c,"D,2",E3,f,g
CSV

open my $fIn, '<', \$csvFile;

my $csv = Text::CSV->new();

$csv->column_names($csv->getline($fIn));

my $arrayref = $csv->getline_all($fIn);
my %dataHash;

$dataHash{join "\t", @{$_}[0, 1, 2, 5, 6]}{$_->[3]}{$_->[4]} = 1 for @
+$arrayref;

for my $key (sort keys %dataHash) {
    my @fields = split "\t", $key;
    my @dKeys = sort keys %{$dataHash{$key}};
    my %eKeys;

    $eKeys{$_} = 1 for map {keys %{$dataHash{$key}{$_}}} @dKeys;
    
    my $dField = join '; ', @dKeys;
    my $eField = join '; ', sort keys %eKeys;

    splice @fields, 3, 0, ($dField, $eField);
    $csv->say(*STDOUT, \@fields);
}
[download]

Prints:

a1,b,c,"D,1; D,2","E1; E2",f,g
a2,b,c,"D,1; D,2",E3,f,g
[download]

Update: small bug to deal with repeated E column values