T0498 and T0499

[guest]

I knew that there is a 2007 PNAS paper talking about two proteins similar to T0498 and T0499.
It is amazing that there are still four servers (from three different groups) that can guess T0498's fold correctly.

[guest]

Here is the link to the paper in case some one wants to read it.
http://www.pnas.org/content/104/29/11963.abstract

I knew that there is a 2007 PNAS paper talking about two proteins similar to T0498 and T0499.
It is amazing that there are still four servers (from three different groups) that can guess T0498's fold correctly.

[Guest]

Structures of these two proteins have been clearly plotted in this
paper which provide plenty of information to model both proteins.
This may be unfair for those predictors who did not read this paper.
Therefore, these two targets should be removed from the final evaluation.

I am wondering whether there are other targets like this...

[Guest]

Structures of these two proteins have been clearly plotted in this
paper which provide plenty of information to model both proteins.
This may be unfair for those predictors who did not read this paper.
Therefore, these two targets should be removed from the final evaluation.


Maybe the original poster knows something the rest of us do not, but there have been
several other successful redesign experiments on Protein G - all of which involve small numbers of
mutations. It was fairly obvious that T0498/T0499 were protein design pairs but most of the work
on this protein has been in designing domain-swapped multimeric variants of the protein. Given
that the structures of T0498 and T0499 have not yet been released, as far as I know, we can't know
for certain which particular fold variant of Protein G this is. Even with that paper in hand, guessing
that T0498 folds like the first structure in the Alexander et al. paper may or may not be correct - it
could just as easily be a domain-swapped variant of the wild type protein - or maybe something else
entirely!

Nevertheless, I do think it would be more useful for these CASP "trick questions" (there have been others
in previous CASPs) to be posed a little more clearly. I guess it's part of the challenge - but more people
might have attempted a non-template-based prediction if it was clearly stated that these proteins
had been designed to have different folds, and everyone would have know to look at the literature more
closely.

[guest]

I think it is totally fine to use T0498 and T0499 as human targets, but it is not very fair to use them as
server targets.

Structures of these two proteins have been clearly plotted in this
paper which provide plenty of information to model both proteins.
This may be unfair for those predictors who did not read this paper.
Therefore, these two targets should be removed from the final evaluation.


Maybe the original poster knows something the rest of us do not, but there have been
several other successful redesign experiments on Protein G - all of which involve small numbers of
mutations. It was fairly obvious that T0498/T0499 were protein design pairs but most of the work
on this protein has been in designing domain-swapped multimeric variants of the protein. Given
that the structures of T0498 and T0499 have not yet been released, as far as I know, we can't know
for certain which particular fold variant of Protein G this is. Even with that paper in hand, guessing
that T0498 folds like the first structure in the Alexander et al. paper may or may not be correct - it
could just as easily be a domain-swapped variant of the wild type protein - or maybe something else
entirely!

Nevertheless, I do think it would be more useful for these CASP "trick questions" (there have been others
in previous CASPs) to be posed a little more clearly. I guess it's part of the challenge - but more people
might have attempted a non-template-based prediction if it was clearly stated that these proteins
had been designed to have different folds, and everyone would have know to look at the literature more
closely.

[Guest]

I think it is totally fine to use T0498 and T0499 as human targets, but it is not very fair to use them as
server targets.

See your point, but in theory a really clever server could have selected the correct template (assuming the particular
fold is in PDB) or worked out that the mutations were incompatible with the "obvious" fold. I suspect there are no current
servers with that level of sophistication but I see no harm in giving it a try. It'll be up to the assessors whether they include
this data when assessing the server predictions - I suspect it won't matter much either way.

[Guest]

The structures of T0498/T0499 have actually been released (2jws/2jwu, with 2/4 residues
different). The structure of T0498 is exactly the same as the first structure in Figure 2 of
the paper and T0499 as the second.

At the current stage, I do not think (hope I am wrong) there is any algorithm that can
fold T0498 correctly without reading the paper or having other information (given the
fact that there are a number of 'wrong' templates with 60% of sequence identity to this
target). It is fine to put them in CASP. But if the structural pictures are published before
casp and not all the predictors have read the paper during their prediction, it will result
in unfairness here. I suggest T0498 should be dropped off but T0499 may be kept because
at this level of homology and accuracy, the picture does not help anyway.

[Guest]

Last year or earlier in this year I did an experiment on the two proteins described in the paper by running my threading program on them. It turned out that the two different folds appeared among top 20 templates.
However, it is very difficult to have a correct guess of the top 1 models even if I used some energy functions or model quality assessment programs to help me.
T0498 and T0499 are only different in three positions and my program just generated similar folds as their top 1 models.


The structures of T0498/T0499 have actually been released (2jws/2jwu, with 2/4 residues
different). The structure of T0498 is exactly the same as the first structure in Figure 2 of
the paper and T0499 as the second.

At the current stage, I do not think (hope I am wrong) there is any algorithm that can
fold T0498 correctly without reading the paper or having other information (given the
fact that there are a number of 'wrong' templates with 60% of sequence identity to this
target). It is fine to put them in CASP. But if the structural pictures are published before
casp and not all the predictors have read the paper during their prediction, it will result
in unfairness here. I suggest T0498 should be dropped off but T0499 may be kept because
at this level of homology and accuracy, the picture does not help anyway.

[Guest]

The structures of T0498/T0499 have actually been released (2jws/2jwu, with 2/4 residues
different). The structure of T0498 is exactly the same as the first structure in Figure 2 of
the paper and T0499 as the second.


Well, it's no surprise that they have the same structures as those seen in the paper - those ARE the structures
from the paper! They are not T0498 and T0499 however. Maybe T0498 and T0499 do adopt those same structures,
but as it only takes 2-4 residues to change the overall fold of these proteins we can't say for sure until we
see the structure solutions for the exact T0498 and T0499 sequences.

[Guest]

Another paper about these two proteins has been published in PNAS:
http://www.pnas.org/content/105/38/14412.abstract

Both papers were edited by David Baker.