Feature Requests item #1588916, was opened at 2006-11-01 22:56 Message generated for change (Comment added) made by aherraez You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=379136&aid=1588916&group_id=23629
Please note that this message will contain a full copy of the comment thread, including the initial issue submission, for this request, not just the latest update. Category: None Group: None Status: Open Priority: 5 Private: No Submitted By: friedar (friedar) Assigned to: Bob Hanson (hansonr) Summary: alpha carbon smooth trace/ribbon/etc. Initial Comment: Jmol should have an option (and actually IMO it should be the default) of smooth backbone representations that follow the alpha carbon position rather than the current default of the peptide bond. Since the work to accomplish this is not trivial, I'd like to justify the need for this rather strongly. There are two compelling reasons. One is that it is simply the standard in crystallography and NMR representations of proteins, so it is hard to understand why Jmol should execute this very standard representation differently. An alpha carbon trace has a characteristic look and feel that protein structure people expect to see, because it is more reflective of the true path of the backbone. The muted curves of the backbone in Jmol's sheets, turns and loops are actually quite different from the undulations of an alpha carbon trace/ribbon/cartoon. Second, in proteins there is a frequent need to display a few key side chains as sticks (or ball and stick) protruding from a smooth trace or cartoon at an active or binding site. This type of display is used in almost every structure representation of a ligand binding site. With the current Jmol representations, the alpha carbons of these key side chains are hanging in space, unconnected to the backbone tracing. This looks odd and is, of course, incorrect. Using a strict "backbone" display that connects the alpha carbons with straight rods is insufficient, since the straight rods echo the stick nature of the sidechains, and it is harder for the eye to discern the side chains. The smooth representations are important in this context. Thanks for listening. ---------------------------------------------------------------------- Comment By: Angel Herraez (aherraez) Date: 2006-11-02 19:50 Message: Logged In: YES user_id=1065324 Frieda, Indeed, the documentation at Jmol's website http://jmol.sourceforge.net/docs/JmolUserGuide/ch04.html#trace is wrong, claiming that trace uses "the center of the carbonyl carbon-alpha carbon bonds". I see the same in an old copy of Bob's scripting doc, but is correct in the current one. I remember a discussion on this (likely in the users list) and checking the actual behaviour, but cannot find now where it was and where the conclusions were written (clearly on Bob's page at least). I will try to see if I can fix the webpage doc. ---------------------------------------------------------------------- Comment By: friedar (friedar) Date: 2006-11-02 18:25 Message: Logged In: YES user_id=1190355 Examples of programs that use an alpha carbon trace: Molscript, PyMol However, I note that the cartoons in these programs can use a more averaged representation for sheet structure. In the cartoon/ribbon mode, the traces that connect the helices and sheets still go through (or very close to) the alpha carbons, but the sheets do not wave around as much. So I must agree with you, Bob, that the cartoons do smooth over the pleats. My preferred solution would be to have each representation author- configurable, so as to be able to combine an alpha carbon trace with sheets that have less curvature. However I am also interested in discussin the alternative approaches if others think they are advantageous. I am interested to read the excepts you included on how things are rendered in Jmol vs. RasMol. Clearly I was wrong about the peptide bond, and this is because I was incorrectly recalling this Jmol documentation of trace:"A "trace" is a smooth curve through the center of the carbonyl carbon-alpha carbon bonds of a peptide". So I was recalling the incorrect bond, but still, the documentation seemed to say that the trace went through a bond, not the alpha carbon. Maybe this documentation had it wrong anyway, since you comment: "Jmol: the control points are as follows. default: midpoints between alpha carbons... " Regardless of this documentation confusion, the fact that the default trace is more of an "averaged backbone" than a wavy one is something I would like to have control over. ---------------------------------------------------------------------- Comment By: Angel Herraez (aherraez) Date: 2006-11-02 18:00 Message: Logged In: YES user_id=1065324 Thank you, Bob. You are right. And we need a quiet discussion by users on this. > Angel, I am reticent to change a default for such a major > thing as the shape of all cartoon renditions without > significantly more input. We can do it, but it's a MAJOR > change. Not hard to code, but it might be hard on users. We > need to hear from Eric, Tim, and others first. Certainly. I was just giving my opinion. On the other hand, those of us coming from Chime have already had to get used to slightly different renditions when moving to Jmol. > "The RasMol trace command displays a smooth spline between > consecutive alpha carbon positions. This spline does not > pass exactly through the alpha carbon position of each > residue, but follows the same path as ribbons, strands and > cartoons." Yes. The problem is, they don't define precisely what is the path under "ribbons" heading. > RasMol: the control points are midpoints between alpha > carbons EXCEPT FOR HELIXES, where a special 1-Angstrom > displacement toward the outside of the helix is applied in > order to make it bigger. This is a valuable piece of information, which I didn't have (it's not in the manual). However, there must be a difference since Jmol and Rasmol renderings for sheets seem not to coincide: alpha C are closer to the trace in Rasmol (raising Frieda's concerns about sidechains hanging out from nowhere). It looks to me as if in Rasmol the spline is attached to midpoints, but closer to the alpha carbons. Is there a weighting factor for the curvature? > RasMol: The "curve" is not actually smooth. It is > actually a > set of four straight segments drawn between control points > and three calculated midpoints. > (http://www.bernstein-plus-sons.com/software/rasmol/src/repres.c) Yes, I saw that before by using "trace 1" and "spacefill 40" for Calpha. Better work with Jmol, as could be expected ;-) > Now, we can talk about how to change Jmol's cartoons to be > more or less wavy, to be more or less like RasMol, and while > it's not a big deal to change it --- we just define > slightly > different set of center points --- it is important, I think, > to consider what people REALLY want to see. One possibility > is that the curvature is user-adjustable. I haven't got a strong position as to what I prefer, but as I said, too many options won't be good on the user. We need a consistent rendering, not so flexible that it hampers recognition by the outside user. Think of automatic renderings on databases. In general, I think closer to Rasmol is good; we have wonderful new capabilities in Jmol, but the common ones are best kept similar. > 3) Use a combination of alpha and mid-alpha points, > depending upon the structure, similar in a sense to what > RasMol does. This sounds promising to me. It could be the same as giving more weight to the alpha carbons, so that while the trace is fixed on midpoints, it more closely approaches the alphaC. ---------------------------------------------------------------------- Comment By: Bob Hanson (hansonr) Date: 2006-11-02 15:30 Message: Logged In: YES user_id=1082841 Frieda, I would very much like to have concrete examples of what you are talking about in relation to other programs -- I haven't been able to find any on my own that do what you suggest professionals expect to see. I know for a fact that RasMol does NOT do what you suggest. I agree with Angel that pleated is, after all, pleated, and there is merit to showing ribbons more wavy. BUT I could also say that molecular orbitals are wide, and maybe I should show lcaoCartoons in a different way. The point is, a cartoon is just that -- a cartoon. No more. It is an artistic rendition attempting to show some aspect of struture. The standard as far as I can tell is to smooth over the pleats and not accentuate them as would happen going strictly from alpha carbon to alpha carbon, as the current traceAlpha does. Angel, I am reticent to change a default for such a major thing as the shape of all cartoon renditions without significantly more input. We can do it, but it's a MAJOR change. Not hard to code, but it might be hard on users. We need to hear from Eric, Tim, and others first. Frieda, as for "following the peptide bond" -- actually, Jmol does not "follow the peptide bond" -- what it does is only use the alpha carbons, like RasMol. So Jmol already "follows the alpha carbon positions" as you state you desire. (You may be referring to what "backbone" does, but that is altogether different.) FYI: From http://www.bernstein-plus-sons.com/software/rasmol/src/rasmol.1.html "The RasMol trace command displays a smooth spline between consecutive alpha carbon positions. This spline does not pass exactly through the alpha carbon position of each residue, but follows the same path as ribbons, strands and cartoons." Both RasMol and Jmol use a Cardinal Hermite Spline, which is guaranteed to go directly through certain "control points" (http://www.cubic.org/docs/hermite.htm, http://en.wikipedia.org/wiki/Hermite_curve), which are defined in terms of alpha carbons. The differences between Jmol and RasMol are as follows: RasMol: the control points are midpoints between alpha carbons EXCEPT FOR HELIXES, where a special 1-Angstrom displacement toward the outside of the helix is applied in order to make it bigger. Jmol: the control points are as follows. default: midpoints between alpha carbons; set traceAlpha: the alpha carbons themselves. RasMol: The "curve" is not actually smooth. It is actually a set of four straight segments drawn between control points and three calculated midpoints. (http://www.bernstein-plus-sons.com/software/rasmol/src/repres.c) Jmol: The curve is truly smooth -- the number of midpoints depends upon the zoom level and the screen resolution, varying anywhere from about 5 points to several hundred per segment. Now, we can talk about how to change Jmol's cartoons to be more or less wavy, to be more or less like RasMol, and while it's not a big deal to change it --- we just define slightly different set of center points --- it is important, I think, to consider what people REALLY want to see. One possibility is that the curvature is user-adjustable. I think I could write a method to do this. Possibilities include: 1) Leave it like it is. 2) Just use alpha carbons. Don't worry about what other programs use. 3) Use a combination of alpha and mid-alpha points, depending upon the structure, similar in a sense to what RasMol does. 4) Allow some user-defined customizations -- or not. 5) ? Bob ---------------------------------------------------------------------- Comment By: Angel Herraez (aherraez) Date: 2006-11-02 11:37 Message: Logged In: YES user_id=1065324 I agree with Frieda, and I vote for having this as the default (if not the only one; we are getting things complicated with so many nice options; it's great for authors to have the options, but I think the program should not offer two ways of tracing, will be confusing to users). Yes, the wavy sheets don't look too good, but in the end they are *pleated* sheets, right? The backbone is zig-zag, so that's what we (must) see. I checked Rasmol (2.7.3.1) and the "wavyness" is there, more than with Jmol's set traceAlpha OFF but less than with set traceAlpha ON; the trace/ribbon does not go right through the center oh the alphaC atoms, but slightly to one side, making very little distance in helices but noticeable in sheets and turns. Maybe someone can check the code used in Rasmol. ---------------------------------------------------------------------- Comment By: Bob Hanson (hansonr) Date: 2006-11-02 05:30 Message: Logged In: YES user_id=1082841 done now for all secondary renderings: cartoons, meshRibbons, ribbons, rockets, strands, trace HOWEVER -- while I agree that helices look terrific with an alpha trace, I'm not sure you will say the same about beta-pleated sheets. They are awefully wavy. I notice that RasMol does not do this. Kinemage does not do this. (Protein Workshop's sheet representation is pretty messed up, so let's not compare that.) I suspect that you do not really want it, either. What do you really want for sheets? Bob ---------------------------------------------------------------------- Comment By: friedar (friedar) Date: 2006-11-02 00:43 Message: Logged In: YES user_id=1190355 OK so is this the right form box to use for response? I hope so! 1 - thanks for implementing trace, I look forward to checking it out. 2 - Yes, I think this is needed for cartoons and ribbons. Again, the path of the backbone looks quite different and I think we should be able to display it correctly. Rockets - I am not sure we need it for this, I can't quite imagine the difference in the rocket position, but I suspect it would not be a very big or meaningful difference. I have asked Eric (Martz) for his thoughts on the rockets aspect. ---------------------------------------------------------------------- Comment By: Bob Hanson (hansonr) Date: 2006-11-02 00:07 Message: Logged In: YES user_id=1082841 trace is easy and is now (10.9.90) enabled by set traceAlpha which can be toggled back and forth to see the difference. Are you saying that you need this as well for cartoons? I can see what it is about in terms of the trace; what would be implied in terms of rockets or cartoons? Bob ---------------------------------------------------------------------- You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=379136&aid=1588916&group_id=23629 ------------------------------------------------------------------------- Using Tomcat but need to do more? Need to support web services, security? Get stuff done quickly with pre-integrated technology to make your job easier Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642 _______________________________________________ Jmol-developers mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/jmol-developers
