I've read your question four times now, and I still don't understand what you're trying to do. It almost sounds like you want to match approximately (even nondeterministically?), a purpose for which regexes aren't suited. I'm also unsure of the purpose of your sample regexes: /atgc/ will match only that exact sequence, not combinations of the member letters. What are you matching with those, nucleotides or sequences? I guess I'm just generally confused what the "weighting" will be applied to, and generally what your goal is.