The Alignment Tool: Purpose and Prospects

[beta_helix]

During protein evolution, amino acid sequences generally change faster than three dimensional structures. Evolutionarily related proteins almost always have similar structures, and hence to predict the structure of a protein it is very useful to identify proteins with similar amino acid sequences whose structures are known. The sequence of the protein of interest can then be "threaded" onto the known structure, which is referred to as the template. The resulting structure is a good starting point for additional refinement.

While automated methods can correctly choose the right template structure for very closely related proteins, when the sequences have diverged considerably during evolution identifying the correct template and aligning the sequences correctly (such that corresponding residues in the three dimensional structures are properly superimposed) can be quite difficult. Starting from the wrong template and/or the wrong alignment puts you that much further from the best scoring structure.

This is where the Alignment Tool comes into play. The Alignment Tool allows the user to hand thread a sequence over potential homologues, giving human spatial reasoning a chance to do what automated methods frequently can't. In addition, the alignment tool allows users to mix and match the best parts of each alignment into a hybrid template since different sequence alignment techniques have their strengths and weaknesses.

Bottom Line: In this year’s CASP, the new alignment tool offers players a potential edge over everyone else.

[brow42]

I guess everybody is too baffled to think of any! I think the new alignment tool is very exciting.

I may not make the chat so here are my comments /questions.

Visualization

Visualization in general is hard…it's hard to see the protein and hard to see the template. It's hard to find which part of the primary sequence corresponds to which part of the folded image.

Eventually, what's worked best for me is to select large sections with the partial threading tool, see where the flashing is, and search for the region I'm interested in. This takes a lot of mouse clicks. Just moving the cursor along doesn't work well since the flashing is invisible in a big protein. Often, I try to go by segment number, but the tab popup is disabled while the alignment tool is up…I have to exit, tab, align. And this only works for the protein, not for the template sequence.

Cutpoints

It's hard to align / predict how a cutpoint will score when it is closed out. With the alignment tool, it's utterly trivial to get a cis peptide bond, since it's just a translation of one unit. I suppose I should be looking closer at the sidechains. Rebuilding doesn't always fix these, and also, wiggle can get stuck when there's really really bad clashing. Basically, undo and try reconnecting again.

I apparently got one of these cis-bonds and didn't notice it right away…after I'd fixed all of the 'red' segments everything looked fine…except for this one non-bonding sheet that had distributed the flip, during all the wiggling, into a twist along the entire strand. I couldn't fix this…rebuilding and tweak just made it all blow up, and the big sidechains prevented gentle movement by banding. I had to abandon it all and rethread.

This also makes any future scissors tool not as useful as it could be. I'd be afraid to use it it most cases.

Saving

Alignments are global to all tracks, and I think, are reset on puzzle load. If I realize I need to make a change to a fold after I've gone off to try other alignments, or restart the client, I've lost the alignment, and I can't make a small change. There's also no way to communicate alignments to teammates.

Cannot Thread

No idea what causes this message. I understood that before cutpoints, it would not thread if it would strain the peptide bond. Now, I'm not so sure. And, I have no idea how to fix it or where the problem section is, unless it is the most recent change I just made. If not, I have to reset, and all work is lost.

Also, whatever is causing it, it demands immediate attention from me…since it won't thread, I can't work on another part of the sequence. Perhaps for the best if I'm going to end up resetting anyways.

In the recent quest to the alignment 546, template seg 57 goes GV..N.V…….V (N is an insertion, . are deletions relative to the protein). The threading tool rejected almost any change I made to that region. I've no idea why. I'm also almost convinced there's a memory effect in there somewhere.

General

The threading toggle button either doesn't change, or changes imperceptibly (I'm on a 1920x1600 LCD). Never really quite sure what state its in, unless threading is on and I move something.

Like a lot of other newcomers, I have no idea what makes a good fold, so I have no idea how to use the alignment tool. In the end, all I did in 546 was make some unit shifts to flip the sheets…and I knew to do this only because of the guide.

Others have noted that a partially aligned segment may be buried and must be extricated, which can be difficult if not impossible.

[brow42]

Part of what makes the partial thread tool difficult to use is the difficulty to enlarge or shrink the existing selection. The selection must be unlocked, changed, then locked, and its easy to lose the selection entirely. Also, I often find my self wanting to select on the template, but selection is only possible on the protein.

Until a few days ago, I had no idea the score was dynamically calculated from the current alignment. I thought it was essentially a precomputed number from the server!

[bertro]

Some questions and thoughts,

1- As brow42 said, very hard to see what you are doing in a big protein. Partial threading is helping sometimes as it reduces the field of view.

2- I would like to see a way to preserve my work. Save it with the track. Transfer it between tracks.

3- And also a way to go from the protein current shape back to a 'close' alignment (I guess it would be the reverse of threading)

4- Never reset the alignment, unless I press a reset button.

5- When I am working with the top row, it is very difficult to control the left-right movement. It seem to have a life of its own. The same with a template, I move it and the top goes bye-bye. Frustrating.

6- What is the meaning of the score value at the right of a template? It does not seem to correlate with the white bars. I can have lots of big white bars and not a good fold!

7- Should I concentrate on the spatial arrangement instead of / and the score?

8- Someone created this tool. I would like to have this person explain how he/she envisioned its use. Maybe take a 'small' known protein and a not obvious template (like the third one in recent puzzles) and make a video showing the tool in action. Try to explain the reasons for setting up the template and protein in such and such a way.

[alwen]

My comments:

As already noted, on the dark background it's hard to see what the tool is doing.

The "lock/unlock" button is not intuitive as a way to make the alignment "take". I know what it does now, but I kept looking for the old "Go" button.

Plus you can't see the difference (one or two gray pixels against a white background) between the locked versus unlocked state of the button. It's too small and doesn't have enough contrast.

And ugh! I have the same problem as bertro's 5 - I start pushing a sequence right or left and it moves as if it were greased.

[Rav3n_pl]

• we can lock upper and/or lower parts to prevent form moving
• threaded molecule and template guide are visible like normal protein (not so transparent like it is now)
• we can select ON PROTEIN parts and quick jump to that part in tool
• we can select on protein part we want to change (ie select part of sheet, move it aside in tool to flip)
• we can see sidechains on protein in stubs mode like in qttn
• we have kind of undo graph for tool
• we can save/load aligments to tool

• we can make ONE thread from MULTIPLE templates simply selecting what parts we want from what template

• add scissors tool :D

[spmm]

that would let people copy really easily

[B_2]

I found the old alignment tool baffling, and the new one is just ever so much more incomprehensible to a layperson (non-scientist).

I have no idea what any of it means, and just end up mashing things around until the restrictions are removed, and now there is the added annoyance of trying to get those blue stretchy things to go away.

[brow42]

<silverberg: I have been fiddling with 550, and looking at the alignments makes me slightly irked.
Brow42irc: I'm kinda awake so I can answer some questions about it
silverberg: As in, "what's the best alignment?"
silverberg: ta brow
Brow42irc: well that's a hard one :)
harp: or how do you know when you have reached the best alignment
silverberg: yep, it is.
Brow42irc: I can tell you what bad alignments are
harp: go on…
silverberg: It does feel a bit hit or miss, for me.
Brow42irc: so, first you want to turn on stubs and hydro coloring…if a sheet has hydrophobics sticking up, then it needs to be flipped over
Brow42irc: instead of tweaking, you can use the alignment tool and shift it sideways by 1, which flips it
Brow42irc: prolines and glycines usually show up at the end of sheets and helices, so you want to put them where they show up in the template….or if not, then arrange that they end up in a u-turn
Brow42irc: if you have a proline in the middle of a sheet or helix, that's probably a bad alignement
Brow42irc: an odd-length gap in the alignment tool will create a 180 degree flip, which is usually bad, you will have to rebuild it out, and that flip will have to be spread out over many residues…try not to have those.
Brow42irc: Space over Space is meaningless, that's why when you move one line, the other line moves instead
Psych0Active: The options in the alignment tool are kinda cryptic
Brow42irc: Space on one side, and space on other side, offest by one, is also meaningless…close that up into a simple substution (spaces mean insert and delete)
Psych0ActiveIRC: Ah! :)
Psych0ActiveIRC: Indeed! Thanks!
Brow42irc: score is what determines the so-called 'best' alignment which is what the computer used. Something like exact match is worth, say, 3, and type match is worth one, and mismatch is worth -1 and space is worth -2
Brow42irc: however, the computer isn't looking at the actual shape of the template, just the letter sequence…that's why you cn modify the computer's guess to find a better alignement
Psych0ActiveIRC: That's starting to make sense, especially about gaps
Brow42irc: yes. so a space in the template is an INSERTION of extra AAs from the real protein
Psych0ActiveIRC: k
Brow42irc: a space in the protein is a DELETION of AAs from the template
silver: I'd like to know what to look for when I first open the Alignment Tool
silver: All these colours, letters and spaces, and stuff.
Brow42irc: okay, if the scientists give my a SS guess, I apply that first (remember to use ctrl shift 9!)
Brow42irc: I then I align to each template. I ignore guesses that don't match the SS guess
Brow42irc: I find a template that seems to be the same structure (sheets are sheets, helices are helicies) and desont' have a lot of ugly cuts or unthreaded parts
Brow42irc: and then I start looking at that template for upside down sheets
Brow42irc: so, that's the first thing I do
Brow42irc: I also switch the template from AA to Type
silver: Ahh, that is useful.
Brow42irc: later on I might look at other templates
Brow42irc: that I skipped or whatnot
Brow42irc: if another template can give shape to an unthreaded part, then I might use partial threading there
Brow42irc: in 550, the SS prediction had that helix at the front, so I immediately rebuild that after threading…then I started looking around to where to put the helix
Brow42irc: here I used the alignment tool to cut off the helix:
Brow42irc: http://fold.it/portal/files/chatimg/irc_269484_1335968894.png
Brow42irc: I brought up an unused template the make the spaces not go away
Brow42irc: I should say, before I stop threading, I check all the uturns
harp: for what exactly Brow?
Brow42irc: because of alignment shifts either by me or the computer, glycines and prolines that should be in the uturn may have moved into straighter loops or sheets or helices, so I shift them back to the uturn.
Brow42irc: I may also use that short segment to change an odd-odd gap around it to even-even (basically flipping the short segment)
Brow42irc: shifts of 2 don't cause a flip, so if a sheet looks fine but the uturn is messed up, I might shift the uturn and sheet 2 spaces.
Brow42irc: for navigating the alignment tool, if I can't see the flashing, I temporarily make a long segment, which flashes together
Brow42irc: I mean, long selection
harp: That is probably not the end of this tut, but that is a lot of information to absorb and useful, brain hurts now, but thank you for it
Brow42irc: yw

[Susume]

I would find it very useful to have the secondary structure of each template residue showing in the alignment tool. I'd also like to see in the alignment tool the secondary structure I have assigned to each residue in my current model. I am much more likely to scoot something over in the alignment tool on the basis of "this should/should not be part of the helix" than on the basis of "gee, these are both purple." It would also make it easier for us to place cutpoints in the loops rather than in sheets or helices. And it would make it easier to see which residues in the alignment tool match which residues in the picture.