Diderot's Suggested Method Edit

Note: This page is under construction and will change suddenly and rapidly over the next few days.


First, try to develop an intuitive sense of what folded proteins look like. mostly lacks the computational power of the distributed-computing approach to the problem. It hopes to make up for that lack with plain old human intuition. This is not as far-fetched as it sounds. In the game of go, for example, even moderately skilled humans still routinely beat the best computers -- because there is just too much math for the computers to do. Humans skip the math and win by intuition.

Protein folding appears to be a similarly complex problem. To develop your intuition, you should spend a lot of time looking at successfully solved protein folds. Ubiquitin is a work of art, as is insulin. Scientific journals, the protein data bank, Rosetta@Home and even Wikipedia are excellent sources for these successful folds. Learn from them.

Strive for an aesthetically pleasing shape. Beta sheets flow. They are virtually never in perfectly straight lines. (If they were, the computational math would be a lot easier, and might not be necessary.) Proteins are not like Le Corbusier's "efficient" buildings. They're not square, straight, and blocky. They're more like Frank Gehry's twisty and curvaceous architecture. And yet there is a logic and a beauty to that flow. A folder is an artist.

This aesthetic sense is what you're going to aim at in every puzzle. From what I've been able to tell, coming up with a good general idea of what a given protein ought to look like is the single most important thing you can do. Having a bad general shape and then making minor adjustments to it may raise your score incrementally, but to really break through requires a sense of the large, important stuff. It requires a sense of the way proteins move, and of the way they look in their finished form.

While You're FoldingEdit

Step One: Alignment and Secondary StructureEdit

The Alignment Tool is very powerful, and I wish we'd had it back when we were doing CASP 8. (Can you imagine CASP without alignment?) This tool should be where you start on every single protein that doesn't already come with a suggested alignment.

That said, do not be a slave to the alignment tool. Even a "successful" alignment is basically a piece of crap. It usually scores deep in the negative tens of thousands, and many, many elements of the protein are likely to be flat-out wrong -- wrong folding, wrong Secondary Structure, wrong Sidechain Position, wrong everything.

Once you get an alignment that is acceptable to the program, look at the protein's secondary structures -- its loops, helices, and sheets. Don't even try to fold the whole thing globally yet. Instead, go through the entire protein, amino by amino, and make sure you feel comfortable with the secondary structure that the Alignment Tool has assigned to each one.

For example, if you find a Proline or a Glycine in the middle of a Beta Sheet or an Alpha Helix, be suspicious. They are probably not rightly assigned. Consider making them into part of a Loop.

If you find several aminos from the MALEK group in the same vicinity, consider making them into a helix. Your helix could include aminos not in the MALEK group, but most of the time a helix will be noticeably heavier on these five.

If you see a bunch of MALEKs roughly near each other, then the chances are that they want to be a helix. Assign them that structure -- including any aminos in between your MALEKs -- and then use the Rebuild tool on that area alone, plus a few segments from the surrounding loops. It will gradually turn this section into a more or less well-formed helix. If there are still deformities, consider that perhaps you have two or more helices, not just one.

For most alignments generated by the client, some aminos will be in beta sheets. These aren't terribly difficult to get into good form, as long as you don't have any prolines or glycines in them. Figuring out which aminos need to be in beta sheets is actually much harder, because almost all aminos can do it. A lot of this work will come in step two, and you might often change your mind about which aminos go in and which go out of your betas.

Step Two: Imagining the TertiaryEdit

At this point your alignment will still look like crap and still score very low. Don't worry about it.

Take a bunch of Bands and attach them to the Backbone so that they pull everything a bit apart -- you want a better look at the whole thing. Wiggle it, and let the bands do their work. Take note of how things seem inclined to move. See which parts of the protein might make good beta sheets, and how these might be brought around to hydrogen bond with one another.

Remember that at this point you are called on to be the most creative and the most thoughtful. This is likely what makes or breaks your fold.

Take your betas and get them to hydrogen bond with one another. You may need to rebuild some of them, and you will almost certainly Tweak them a lot. Consider which sides of these betas are hydrophilic or hydrophobic. A beta sheet with a very hydrophilic side will almost certainly be placed on the outside of the protein, with the hydrophilic side pointing out. Hydrophobic sides of your betas will likely face the inside. A beta sheet that is hydrophobic on both sides will probably lie on the inside of the protein, surrounded by helices and loops, or by other betas.

Drag things around by brute force, making liberal use of rubber bands, wiggles, and tweaks. Only rarely shake the sidechains, and avoid Local wiggles entirely. At this point you don't want to get too much Mojo. You want to be able to stay flexible and implement your big structural ideas easily. You still want to explore what's out there, and see what possibilities there are -- not to get caught in a local minimum.

When you are satisfied that you have a roughly good framework, remove your rubber bands, unfreeze any sections you've frozen, and wiggle and shake the whole thing into place. If you've done well, your score will rise dramatically, and your protein will stabilize at something like what you meant it to look like.

If it doesn't look anything like what you imagined, take stock of the situation. It sometimes happens that you get an even better configuration than the one you imagined. And it sometimes happens that you end up with worse. This is where you need to use your intuition again, possibly go back to the drawing board, consider a new alignment, or just consider some rebuilding to make your first idea work. Much judgment is required at this stage as well. Repeat it until you are satisfied and you have a stable, compact overall shape.

Step Three: Unhappy Aminos and Other Fixable ProblemsEdit

One of the first things you'll notice is that many amino acids are still low-scoring, "unhappy" parts of the fold. You'll want to address these next. It could be that they don't really belong in the secondary structure you've given them. Or it could be that the structure itself needs to be rebuilt. Or you might not even be able to say quite what's wrong with them.

There are a variety of ways to deal with this, but I find that small, highly constrained rebuilds are often the best. There are a nearly infinite number of ways of dealing with these aminos, but the fact is, now is the time to do it. Be creative and go for it.

Hydrophobicity will likely remain a problem at this stage, and you will need to address it as well. Often a helix has a hydrophobic side and a hydrophilic side, and these work just like they do for sheets. Rotate your helices to get them to where they belong based on hydrophobicity.

New structural ideas may emerge. Don't be afraid to rethink your structure.

Mojo can often be your friend here, because at this stage it will allow you to keep those parts of the fold that are worth keeping -- work in one local area is less likely to ruin work done in other local areas. But as you continue to adjust, mojo will rise, and fewer and fewer changes will be effective.

After every change in this step, you will want to shake the sidechains, either before or after you wiggle the change into place. (I usually try both.) Sometimes, you'll want to move a sidechain manually and then wiggle, especially if you are trying to fill a Void. By the end of this stage, if you've done everything right, voids will be relatively few, and a well-folded protein will often have no or almost no voids within it.

Step Four: Finishing with ScriptsEdit

Once you've fixed your lower-scoring aminos, your protein should be mostly green. It's at this point that you will want to consider local wiggle strategy, various tweaking scripts (I like Pletsch's Acid Tweaker), and other Endgame scripting tools. Each will raise your mojo to the point where relatively few of your conscious changes will result in score increases, so you do want to be careful here. Do not script too soon, and do not over-script. One thing I like about Pletsch's Acid Tweaker is that it doesn't seem to do this to the protein quite so much as others I've tried.

If You Get StuckEdit

If you become dissatisfied with the fold you've created, do one of two things: Either work on a different puzzle, or go back to the beginning and try a different alignment. Either way, save your work, because you never know when an idea will occur to you about a structure that formerly looked unpromising.