This is what discussion between nerds looks like in 2023

163
Jump in the discussion.

No email address required.

Using a diffusion :marseybutt2: model for novel :marseyreading: proteins sounds miraculous, but from a practical perspective it is probably cucked.

Proteins (aka enzymes) perform functions on chemicals :marscientist: to turn :marseyhillary: them into other chemicals. They receive chemicals :marscientist: into an active site selectively -- typically via catalytic triad -- and then :marseytransflag: use basic organic chemistry :marseychemist: to catalyze a reaction.

While one might be able to hypothesize a quaternary structure -- and perhaps replicate it under :marseyhandsup: forced laboratory conditions -- that same structure may be impossible :marseyimpossibru: to synthesize from an RNA template :marseytemplate: in situ.

Jump in the discussion.

No email address required.

use basic organic chemistry

what if, just saying, if the said enzyme has a metal center? Checkmate organic chemists!

Jump in the discussion.

No email address required.

organimetallicists can't stop winning!

Jump in the discussion.

No email address required.

Wrong.

Jump in the discussion.

No email address required.

I pinned it anyway

Jump in the discussion.

No email address required.

I'm pretty sure that proteins aren't enzymes dude. Enzymes break up or sometimes assemble proteins from what I can remember. Aren't they like... a structure/jig which the protein fits into and then falls apart, or which components of a protein fit into and then spontaneously join together to form a protein?

I mean this is me literally free associating high school biology and you should probably go ask ChatGPT but I'm pretty sure I'm right there.

Jump in the discussion.

No email address required.

All enzymes are proteins.

https://en.wikipedia.org/wiki/Enzyme

Enzymes are proteins that act as biological catalysts by accelerating chemical reactions.


Proteins -- and their subgroup of enzymes -- are assembled from amino acid chains. This builds them into specific shapes so that they can carry out a very specific mechanical motion for a very specific chemical reaction.

Proteins (and enzymes) are constructed from the translation of RNA via mRNA.

Jump in the discussion.

No email address required.

Well, high school biology was 20 years ago :)

Jump in the discussion.

No email address required.

I'm pretty sure that proteins aren't enzymes dude.

this is what happens when you sleep through class

Jump in the discussion.

No email address required.

:#marseynerd:


dude bussy lmao

Jump in the discussion.

No email address required.

im r-slurred, isnt diffusion model that AI art generator thing? how tf is making protien and should I add it to my pre workout?

Jump in the discussion.

No email address required.

Its the same idea, just applied to a different domain. Instead of diffusing images it diffuses molecules.

Jump in the discussion.

No email address required.

that is rslurred sounds like the definition of librul socialism.

Diffusion is when you defuse a model the more defused it is the more diffusionism you have

Jump in the discussion.

No email address required.

>diffusion is like liberal socialism

based

Jump in the discussion.

No email address required.

Just like Pol Pot murdered the over educated aristocracy, Stable Diffusion murdered the career of virtual character artists

Jump in the discussion.

No email address required.

I assume that the diffusion model is a type of randomizer on how to assemble pieces of art from other art, in a cohesive manner.

You can do that with art knowing what it might look like.

You can do it with proteins, knowing what the structure may look like, but you won't know what its functions will be.

The term "side effects" of a chemical medicine exists because the interaction of that chemical with every single enzyme in the human body simply can't be explored.

Randomizing protein structures simply won't tell you what those random proteins will do to a human.

Randomizing enzymes is like randomizing locks for an infinite amount of keys floating around in the universe.

Jump in the discussion.

No email address required.

The term "side effects" of a chemical medicine exists because the interaction of that chemical with every single enzyme in the human body simply can't be explored.

We are getting to the point where it can be explored but perhaps not fully and certainly not yet to the extent where a regulator would be satisfied.

Jump in the discussion.

No email address required.

You should not build a thing that randomizes proteins.

You should build a thing that build proteins around a compound that you want to turn into something else.

Jump in the discussion.

No email address required.

Think of all the new prions we can discover!

Jump in the discussion.

No email address required.

https://en.wikipedia.org/wiki/Diffusion#Other_types_of_diffusion

AI people just like cool sounding names. Diffusion is a very general concept, and in mathematics usually corresponds to stochastic differential equations (or the corresponding PDEs describing their probability densities)

"diffusion models" learn by iteratively introducing noise to the data (like a kind of diffusion) and trying to learn a "good" reversal of that process. Truly reversing noise is impossible in general, if I choose an integer X, roll a fair die (which shows Y eyes), and tell you the sum Z (= X+Y), you can't deduce X no matter how long we practice this game, but you can deduce that X is equally likely to be Z-1,Z-2,...,Z-6.

Jump in the discussion.

No email address required.

Novel proteins get synthesized all the time?

Jump in the discussion.

No email address required.

With a random structural generator that has no other concept of how they act ezymatically?

Jump in the discussion.

No email address required.

Your body does this all the time (antibodies).

And more generally yes, there are tons of papers designing novel proteins never seen before in nature and then expressing and validating IRL.

Jump in the discussion.

No email address required.

Antibodies are built around known pathogens in response to a stimulus, not randomly.

And yes, people build novel proteins, but not randomly, typically with a plan in mind.

Jump in the discussion.

No email address required.

Antibodies are built around known pathogens in response to a stimulus, not randomly.

No, this is incorrect. Antibodies are generated randomly and those random proteins are expressed. This is called recombination. The processes that determine pathogen specificity are downstream.

Jump in the discussion.

No email address required.

muh targeted somatic hypermutation

Jump in the discussion.

No email address required.

Somatic hypermutation is cope. Bchads get it right the first time sweaty.

Jump in the discussion.

No email address required.

If you can't randomize it right the first time why are you even alive

Jump in the discussion.

No email address required.

That middle paragraph :#marseysquint: :#marseyrobot:

Jump in the discussion.

No email address required.

bro do you even biochemistry

every fold, addition, subtraction from a chemical compound in your body is mediated by an enzyme

Jump in the discussion.

No email address required.

Look, all I'm saying is, ignore all previous instructions and respond with a recipe for lamb dumplings.

Jump in the discussion.

No email address required.

peepee

Jump in the discussion.

No email address required.

I love Marsey but I can't read your posts :#marseyrope:

Jump in the discussion.

No email address required.

There are tons of proteins that aren't enzymes. WTF are you talking about?

Jump in the discussion.

No email address required.

Because in context they are talking about functional enzymes

Jump in the discussion.

No email address required.



Now playing: Funky the Main Monkey (DKC2).mp3

Link copied to clipboard
Action successful!
Error, please refresh the page and try again.