Protein function often depends on protein dynamics. To design proteins that function like natural ones, how do we predict their dynamics? @hkws.bsky.social and I are thrilled to share the first big, experimental datasets on protein dynamics and our new model: Dyna-1! 🧵 - ThreadSky

ginaelnesr.bsky.social • 71 days ago

Protein function often depends on protein dynamics. To design proteins that function like natural ones, how do we predict their dynamics?

@hkws.bsky.social and I are thrilled to share the first big, experimental datasets on protein dynamics and our new model: Dyna-1!

🧵

Comments

grocklin.bsky.social•71 days ago

Very cool!! I wonder what aspects of the structure influence these motions? An additional baseline with a small amount of structure would be very interesting - something like your OHE baseline supplemented with # of 3d contacts at each amino acid, or # of contacts + DSSP for each amino acid?

ginaelnesr.bsky.social•71 days ago

Thanks Gabe!! That's a good question! For sure, there are a bunch of other baselines one could try out, including those. The # of contacts is an interesting way to represent structure in this case... but I presume the # likely changes esp during proteins' motions (or conformational change)

grocklin.bsky.social•71 days ago

Thinking about it more, I guess your AF2 Pair Rep model probably sets an upper limit on how well that simple model would do. But it’s pretty good! Basically wondering how close you could get with something structural but very very simple, like an elastic network model.

grocklin.bsky.social•71 days ago

Though if I understand correctly, you’re identifying things with intermediate flexibility… not too high and not too low.

hkws.bsky.social•71 days ago

Hi Gabe! Yeah we thought of af pair (no MSA) as upper limit but what you’re proposing would also be control for pair rep.
To me, the kicker things we want to predict are coordinated motions that have low prob. Saw hints of that with CypA. I don’t think people would think of that as flexibility …

ginaelnesr.bsky.social•71 days ago

1/ We have large-scale exp observables for structures (PDB) and seqs (UniProt). So, what about dynamics?

That's Rex, a measurable from NMR. Rex increases when an atom moves at the µs-ms timescale.

There is no DB for that, so Hannah curated "RelaxDB"

ginaelnesr.bsky.social•71 days ago

2/ From analyzing RelaxDB, we notice that there seem to be evolutionary trends depending on an aa's timescale of motion: aa's with µs-ms motions are more conserved than aa's with no motions (or ps-ns motions)

ginaelnesr.bsky.social•71 days ago

3/ We also saw that AF2 pLDDT is lowest for aa's w/ ps-ns motion, and there was little discrepancy between aa's w/ no motion & µs-ms motion... interesting!

ginaelnesr.bsky.social•71 days ago

4/ But to predict dynamics, we need more data. Well, ms motion can cause peak broadening, so those aa's are missing assignments in the chemical shift data from NMR.

Where do we find these chemical shift datasets? The BMRB!

Hannah curated another labeled dataset: the "mBMRB"

ginaelnesr.bsky.social•71 days ago

5/ The mBMRB has ~10,000 proteins in it! What did we notice? For one, our dataset contained unique information compared to the PDB.

To start, 45% of the proteins had ≥1 NMR structure but no X-ray/EM structures

ginaelnesr.bsky.social•71 days ago

6/ And for the proteins that also had X-ray/EM structures, we notice that aa's missing chemical shift assignments are distinct from unresolved aa's in X-ray/EM

...and there is no difference in B-factors between assigned & missing aa's

Comments

Posting Rules

Reply