If you know that the characters in question will always be uppercase letters (or some other particular character set that doesn't include the next + or -), it's fairly easy: capture the digits and letters that follow a + or -, and use substr to drop the correct number of letters off the beginning:

#!/usr/bin/env perl
use Modern::Perl;

my $str = ".,a..A,,C..+4ACGTG.,-2TG,,...,a";
$str =~ s/[+-](\d+)(\w+)/substr $2, $1/ge;
say $str;
[download]