I think that if you wanted to reframe it this way, it's more like like editing a 800MB binary with sed/regexes (with some degree of random side effects injected sometimes)

And no way to really test it without just running it in production!

Move fast and break (biological) stuff

Really curious how far are we from being able to simulate the human body's response to drugs/modification.

We are so ridiculously far that it's not really worth thinking about. We're still at the stage of simulating how molecules interact with each other. To work our way up even to the single-cell organism level would be a historic human achievement.

To simulate an entire human body is computationally mindboggling. The number of cells are in the tens of trillions and we'd need to simulate their responses, not just to the drugs we care about, but doing so while "operating" (feeding, sleeping, etc.) the simulated body.

I have no experience in this area, but. Once you get the molecules and their interactions down wouldn't that make all other types of cells easier to produce?

And at a certain point couldn't you just feed the simulation DNA and watch it grow?

It takes supercomputers days to simulate milliseconds of interactions between a million atoms using classical mechanics (ie Newton's Laws, basically a ball and spring model so not even accounting for quantum effects except through very rough approximations in the spring constants). We don't even have an accurate model for water yet.

The oldest code we have is probably ~65 years.

Our genomes are ~1.2B to 3.5B (sexual reproduction and life at all, respectively) years old.

Now imagine trying to patch that codebase in production...

And sometimes you end up with two files instead of one.

Or two flies instead of one.

This is not a bad analogy. I got my start in computers but moved to biology and spent a long time thinking of gene therapy as a hex editor for patching genomes.

The reality is so much more complicated, though, that even good analogies don't provide us with any path forward in terms of actual changes to make.

What part is more complicated? Ie, is it the editing process is imprecise? Perhaps the editing process is like using sed (as mentioned by other people)?

In simple terms, at least.

You mean the editing? The editing depends on fairly complicated processes that originally evolved (to the extent that one can impute function) to repair drastic problems.

The complicated bit comes in knowing what to change. There are some diseases in which the entire disease is caused by a single mutation and changing that one mutation in the germ line would be sufficient to correct for the disease. However, most situations involve complex interactions with many genes in a way that making a single or small simple change would have many side effects, or cause other serious problems.

It's like editing a 500Gb text file on disk by trying to flip the magnetic bits with a tiny magnet

Wait...is that not how HDD's work already?

Except the hex (or binary) editor isn't quite perfect yet. Improving this is obviously very important for gene editing.

I think the problem is also, that the code which was the base for the binary is also pretty crappy legacy code without encapaulation etc.

Or built with an optimizing compiler. Point being, just like with typical binaries, you can't easily read what the program will be doing; it's not apparent in the structure of data.

and the 5GB binary has been aggressively obfuscated, with dead code, duplications, random jumps and weird coupling all over the place - and you don't really have a hex editor, just regexes...