The problem with models like this is they're built on very little actual training data we can trace back to verifiable protein data. The protein data back, and other sources of training data for stuff like this, has a lot of broken structures in them and "creative liberties" taken to infer a structure from instrument data. It's a very complex process that leaves a lot for interpretation.
On top of that, we don't have a clear understanding on how certain positions (conformations) of a structure affect underlying biological mechanisms.
Yes, these models can predict surprisingly accurate structures and sequences. Do we know if these outputs are biologically useful? Not quite.
This technology is amazing, don't get me wrong, but to the average person they might see this and wonder why we can't go full futurism and solve every pathology with models like these.
We've come a long way, but there's still a very very long way to go.
Nice work! Here is an article you may find helpful if you have not already come across it.[0]. You may also want to consider benchmarking against some non ML methods.[1]
Can someone explain what one might use this model for? As a developer with a casual interest in biology it would be fun to play with but honestly not sure what I would do
It’s a self supervised learning architecture, and it’s pretty much universal. The loss function runs on embeddings, and some other smart architectural choices allover. Worth diving into for a few hours, Yann LeCun gives some interesting talks about it
The problem with models like this is they're built on very little actual training data we can trace back to verifiable protein data. The protein data back, and other sources of training data for stuff like this, has a lot of broken structures in them and "creative liberties" taken to infer a structure from instrument data. It's a very complex process that leaves a lot for interpretation.
On top of that, we don't have a clear understanding on how certain positions (conformations) of a structure affect underlying biological mechanisms.
Yes, these models can predict surprisingly accurate structures and sequences. Do we know if these outputs are biologically useful? Not quite.
This technology is amazing, don't get me wrong, but to the average person they might see this and wonder why we can't go full futurism and solve every pathology with models like these.
We've come a long way, but there's still a very very long way to go.
HN's blindspots never cease to amaze me.
I am a structural biologist working in pharmaceutical design and this type of thing could be wildly useful (if it works).
full article: https://huggingface.co/blog/OpenMed/training-mrna-models-25-...
Nice work! Here is an article you may find helpful if you have not already come across it.[0]. You may also want to consider benchmarking against some non ML methods.[1]
0. https://pubmed.ncbi.nlm.nih.gov/35318324/
1. https://www.nature.com/articles/s41586-023-06127-z
What makes this dataset or problem worth solving compared to other health datasets? Would the results on this task be broadly useful to health?
What other "datasets" are you talking about? How do you "solve a dataset" ?
Can someone explain what one might use this model for? As a developer with a casual interest in biology it would be fun to play with but honestly not sure what I would do
You can get your feet wet with genetic engineering for surprisingly little money.
This guy shows a lot of how it's done: https://www.youtube.com/@thethoughtemporium
Basically you can design/edit/inject custom genes into things and see real results spending on the scale of $100-$1000.
Is there something like this in text/readable format?
My main concern is using fungi. If it ends up in my lungs I'm most likely screwed, right?
Yes, but most students produce their best work while infected.
> In Progress: CodonJEPA
JEPA is going to break the whole industry :D
Can you explain this? I haven't heard of JEPA, and from a quick search it seems to be vision/robotics based?
It’s a self supervised learning architecture, and it’s pretty much universal. The loss function runs on embeddings, and some other smart architectural choices allover. Worth diving into for a few hours, Yann LeCun gives some interesting talks about it
https://openreview.net/pdf?id=BZ5a1r-kVsf
What makes these Domain specific models work when we don’t have good domain models for health care, chemistry, economics and so on
>we don’t have good domain models for health care, chemistry, economics and so on
Who says we don't?
Examples please?
No, it's really simple to search for domain specific models being used "in production" all over the place
I didn’t find a single one that outperforms a general model.
Ok, alphafold.
It’s not a large language model
Distributing the load on this will probably be infinitely more useful than “folding at home”
gray goo of the future
hmmmm seems like some fake hype.