Training mRNA Language Models Across 25 Species for $165
We built an end-to-end protein AI pipeline covering structure prediction, sequence design, and codon optimization. After comparing multiple transformer architectures for codon-level language modeling, CodonRoBERTa-large-v2 emerged as the clear winner with a perplexity of 4.10 and a Spearman CAI correlation of 0.40, significantly outperforming ModernBERT. We then scaled to 25 species, trained 4 production models in 55 GPU-hours, and built a species-conditioned system that no other open-source project offers. Complete results, architectural decisions, and runnable code below.
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.
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
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.
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.
JEPA is going to break the whole industry :D
Who says we don't?