Unlike the "C" language, Perl strings are much different from arrays. In Perl, you usually have a choice of which to use. String solutions are usually easier and almost always execute faster. Here is a string-only solution to your problem.

use strict;
use warnings;

my $DNA = "ATATCCCGATCAGG3TT!GCA\n";
chomp $DNA;
print "The length of the sequence is:\n", length($DNA), "\n";

my $nucleotideDNA = $DNA;
#my $count = $nucleotideDNA =~ tr/ATCG]//c;  # Remove and count invali
+ds
my $count = $nucleotideDNA =~ tr/ATCG//cd;  # Remove and count invalid
+s

my $locations;  # Find location of invalids in original string
$locations  .=  "$-[0], " while ( $DNA =~ /[^ATCG]/g );

print "There are $count non-valid nucleotides at locations:\n$location
+s \n";


OUTPUT:
The length of the sequence is:
21
There are 2 non-valid nucleotides at locations:
14, 17,
[download]

UPDATE: Modified one line of code to correct errors identified by AnomalousMonk (below) Original remains as comment.