Jyunmi

One, predicting how proteins fold from their sequences, filling in missing pieces of protein structures so it has to guess realistic shapes, and then designing binders that fit onto known targets.

By juggling all those jobs, it picks up habits from real chemistry and physics.

It learns, for example, that certain shapes are stable or that some patterns of changes in atoms make good binding surfaces.

The team also uses filters after the generation.

They throw out designs that look like they won't fold or they make ugly clashes between atoms or that they don't seem to interact well with the target.

An important detail is they release the code publicly.

So in principle, any lab with enough computing power and structural data can try similar workflows.

Now,

This is all still very early.

So there's a few things that we have to pay attention to.

So first, this comes as a preprint.

It's not a fully peer-reviewed paper yet.

We've talked in the past how papers coming in from archive and things like that are not peer-reviewed, and then they're changing their CS policy so that only peer-reviewed papers come in because there's this huge influx of AI content

processed, or AI in the process of making and reviewing those papers.

So just highlight that this is a preprint.

That means it hasn't been peer-reviewed.

All right.

And second, binding is only step one in making any medicine.

So Boltzgen won't tell you if it's safe on animals or humans, how long or how it will survive in the body.

And can they make it cheaply and reliable?