Could DNA be represented as "an embedding" in an AI model?

Where each Array element is a numerical value that corresponds to the order of TCAG chemicals on the crossbar of the helix of a DNA strand?

Has anyone modeled many DNA strands in this way?

If it was done, what kinds of problems could we solve using AI - assuming we could model any known DNA, and predict out generations?

-----

I guess there are at least a couple options:

1) Model an entire DNA strand as 1 model, with each gene (section of DNA) being a vector. So you'd have, say, a C. elegans model. You could play with that species in an AI sandbox.

2) Model all of DNA (the way we model all of language) to end up with a "DNA AI" that can produce any kind of strand (perhaps inventing genes that don't exist, but could). So here you have a biology sandbox, where you can invent perhaps new creatures, or predict out future states of humans, to test drugs and other things (I guess?)

What do you think, what's possible?