Will this work? We'll see....
One of the most recent design puzzles posted is a human fibronectin
structure (from PDB: 1FNF). It is an all-beta protein that is commonly used
as a scaffold for making antibody mimetics. The reason for using
fibronectin is that it does not contain the disulfides commonly found in
antibodies, and is stable in reducing conditions. Most importantly they can
be produced in large quantities in E. coli. A collaborator of ours takes
advantage of these properties, using fibronectin to host a variety of loops
on its surface and panning for desirable properties in a selection
experiment. It has been observed, however, that the stability of
fibronectin sometimes is drastically reduced when their loops connecting the
strands are replaced with non-native ones, as in engineering fibronectin to bind a particular target.
A considerable amount of protein (and therefore the diversity one can screen)
can be lost due to this. An ideal scenario would be to have an extremely
stable scaffold for which the stability is independent of the binding loops,
therefore the experiments are really testing the properties of the loops.
Just to draw an analogy, imagine you are testing the sharpness of knifes,
but the handles on them randomly fail so when you cannot drive a particular
blade through an object, you never know whether it was sufficiently sharp.
We put out this fibronectin puzzle to see if players can pack the core more
tightly, and in the end we chose a single structure produced by the FoldIt
community for experimental testing. It was the design done by BootsMcGraw.
It was chosen because the model showed very promising traits, with few
holes in the core and is scored relatively well. The top ranking structures
sometimes have an excessively abundant aromatic residues, so we did not try
those. To validate the design before it goes into production (which is
time-consuming, and costly, by the way), we sent the sequence through our
structure prediction algorithm via BOINC, and interestingly it found a
funnel near the native structure (the starting structure in this case).
When compared to the funnel generated with the native sequence found on the
crystal structure, they are quite similar! (see figure) Our collaborator at USC has
agreed to test this out in the lab. We'll soon know if this first-ever
experimentally tested design from FoldIT has an improved stability.
Design: BootsMcGraw, Texas.
BOINC: Firas Khatib, Baker Lab, UW.
Experiment: Terry Takahashi, Roberts Lab, USC.
new puzzle: finding home
The new puzzle "finding home" features a homing endonuclease recognizing its target DNA sequence. Homing endonucleases are proteins involved in transfer and replication of DNA sequences, in most of the cases their own. They recognize specific sites in the genome and are able to cut them precisely at that point and trigger the insertion of a new copy of their own gene. The surface in contact with the DNA is responsible for the recognition and discrimination of DNA sequences. Currently there is a lot of interest in homing endonucleases as tools for gene therapy, replacing mutated genes involved in diseases with correct copies. The challenge is to change the specificity of homing endonucleases to recognize new DNA sequences.
And this could be the topic of some of the future puzzles.
Selection Mode Feedback
A couple months back, we released an update that allowed you to try out a new interface that we have been working on that was based upon selecting portions of the protein and perform actions on that selection. (Review of that update can be found at http://fold.it/portal/node/986573). We're interested to hear what those who have tried out this new interface think about it, and whether or not it is or could be an improvement over the existing interface.
Please comment with your feedback, or simply vote on this post based on whether you like the new interface or not.( Posted by beenen34 73 1944 | Thu, 10/08/2009 - 22:58 | 4 comments )
Exploration Puzzle 1
We're introducing a new puzzle today. It's also a new type of puzzle: one where you have to explore far from the starting structure. You've had a preview of this in the exploration map, and this map (Exploration Map in the Social tab) will be an important tool for these new puzzles.
The exploration map shows a point for each protein structure you make: whether it's by pulling the structure, wiggling, shaking, or rebuilding. Each point has two values associated with it: the Foldit score and how different the protein is from the starting protein. This map shows the protein structure "landscape": the different types of shapes it can take, and the score Foldit gives each shape. You might be exploring a new region on the map that initially gives you a worse score, but by working hard in this new region, you may find the correct shape and get the highest score.
Many of you have probably noticed this in an intuitive sense: when we look at the top scorers for many puzzles, we see some of the most bold explorers: people who explore new regions, and then work hard to improve the score in these new regions.
In evolver puzzles, you start with a protein with some score. Before you get a ranking in the puzzle, you must improve the score. In these new exploration puzzles, you must change the protein significantly before you can get a ranking. Foldit also remembers the original structure, so your score won't count if you go back to a similar shape to the start. Keep in mind that for these exploration puzzles the red dot on the exploration map will represent the starting structure's score (even though you are unable to reach that score without changing the structure significantly).
Rosetta sometimes has problems exploring this landscape. We want to see if we can learn from human players how to explore more effectively. So be bold, be creative, and have fun!( Posted by Seth Cooper 73 1944 | Sat, 09/12/2009 - 00:08 | 0 comments )