Hm, not entirely sure what will make things faster, but a few ideas:

You probably don't need to store your data into an array, and then go through the array, compute your counters right away when reading the data;

Use a hash of counters : $freq{$_} ++;

Don't do the division for each single letter: compute the counters, and only at the end, do the four needed divisions; or calculate 1/@data only once into a variable, and use the variable;

Using tr// is likely to be faster than a regex to count the letters. For example:

$c = "ACCCTGATTGC";
$d = $c =~ tr/A/A/;
print $d; # prints 2
[download]