Sometimes you think "this looks like a perfect task to use these cool RE features," but then you end up with something that does not look as beautiful as planned:

use strict;
use warnings;
use 5.010;

my $str = '((mouse,rat),(human,chimp))';
my $str2 = '(mouse,(human,chimp))';

my @buffers;

for ($str, $str2) {

    @buffers = ([]);

    m[
        ^
            (
                (?:

                    (\w+)

                    # This serves only to get out fast if we parse
                    # something bad (e.g. where closing parens are
                    # missing)
                    (*PRUNE)

                    (?{
                        push @{$buffers[-1]}, $2;
                    })
                |

                    \(
                        (?{
                            push @{$buffers[-1]}, my $new = [];
                            push @buffers, $new;
                        })

                        (?1)
                        (?{ pop @buffers; })
                    \)
                )

                (?:
                    ,
                    (?1)
                )?

            )
        \z
    ]x ? do {
        use Data::Dumper;
        print Dumper $buffers[0];
    } : do {
        warn "no match: $_\n";
    };
}
[download]

I look forward to suggestions on how to simplify this code ;)