10 20 30 40 50 60
70 80 90 100 110 120
Should we assume this protein forms disulfide bonds or not?
Cys position Distance Bond Score
116 - 118 2 KFISTCACEIV-ISTCACEIVGG 0.01037
116 - 128 12 KFISTCACEIV-GQIVTCAKEIK 0.01037
118 - 128 10 ISTCACEIVGG-GQIVTCAKEIK 0.0111
and from http://disulfind.dsi.unifi.it/monitor.php?query=jKDK6k
DB_state 0 0 0
DB_conf 3 1 2
In the DiANNA part above, do scores near 0.01 mean
that line's disulfide bond is not very likely?
In the bottom part, what does DB_state = 0 mean?
Also, what does DB_conf = 1 2 or 3 mean?
Going to http://clavius.bc.edu/~clotelab/DiANNA/
and clicking on Help! says the following:
For each pair of cysteine in the input sequence,
a neural network trained to recognize disulfide bonds
produce a score ranging from 0 to 1 (higher the score,
higher the prediction reliability)."
It also gives an example with 4 cysteines and
6 possible disulfide bonds with scores ranging
from 0.1 to 0.9. It picks 2 disulfide bonds each
with a score of 0.8 as the best combination.
The link given for the disulfind part above says:
DB_state predicted disulfide bonding state (1=disulfide bonded, 0=not disulfide bonded).
DB_conf confidence of disulfide bonding state prediction (0=low to 9=high).
So I guess it predicts that no disulfide bonds form,
but it is not very confident in this prediction.
Any ideas why this 140 residues section of the NSP2 protein is not buried inside the other 498 residues of NSP2 protein?
We cannot ignore hydrophobicity in Foldit, so best strategy to have top score on this puzzle seems to be skipping any secondary structure prediction and even the real protein SS and build your own.
My understanding is the CASP organizers have tentatively divided the larger viral proteins into smaller sections (domains) based on the predictions they got back in the first round of competition. Those predictions came from both servers and human teams.
Ok, now it became a bit clearer why do we have residues 360-499 of the NSP2 protein. That looks like a domain of the highest distance similarity between different CASP models. And that part has the highest helix propensity according to the SS prediction.
But I still cannot see any evidence that this fragment is spatially separated from the rest of the NSP2 protein to search its highest Foldit score. The Hiding score looks important enough to skip any structure trying to hide hydrophobics of this NSP2 fragment inside itself.
And btw, overall NSP2 has 27 cysteines.
That's right, Serca. We are simply going off of suggestions from the CASP organizers about tentative domain assignments of this target.
These suggestions likely come from inter-residue distance prediction models, similar to AlphaFold. As far as I know, nobody has collected any empirical data about this protein's structure. So, this sequence might form a well-folded domain; but it might not. Foldit predictions might help us figure that out!
I ran my own psipred prediction on this. Only place it varies significantly is that it predicted the small helix to just as likely be a sheet, so I stripped color from any segments that had a low confidence and converted the small helix to a sheet, making a small sheet pair with cysteine at the end of each. Though I haven't been able to get a disulfide bond that keeps the score up, it puts all 3 cysteine together in the same small pocket.