Re: [Jmol-users] Interpolation in animation?
On Tuesday 01 August 2006 21:20, Eric Martz wrote: 1. I would be pleased to see a simple linear interpolation option built into Jmol. The need for such morph animations is becoming more common as more structures are solved in different conformations. Having it built into Jmol would save a lot of time and make this kind of visualization more accessible to a much wider authorship. Thanx for this elaborate discussion, though I haven't read it in detail yet. But I do already want to mention an concept triggered by your email, and might also be mentioned in your email... Some long time ago I was discussion interpolation with someone (might have been Hens...?), and this someone suggested that a linear interpolation is fine, *especially* if done on the internal coordinates instead of the Eucledian ones. Using the internal coordinates ensures that bond length and angles are reasonable for the interpolated steps. Egon -- [EMAIL PROTECTED] Cologne University Bioinformatics Center (CUBIC) Blog: http://chem-bla-ics.blogspot.com/ GPG: 1024D/D6336BA6 - Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT business topics through brief surveys -- and earn cash http://www.techsay.com/default.php?page=join.phpp=sourceforgeCID=DEVDEV ___ Jmol-users mailing list Jmol-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/jmol-users
Re: [Jmol-users] Interpolation in animation?
Egon Willighagen wrote: On Tuesday 01 August 2006 21:20, Eric Martz wrote: 1. I would be pleased to see a simple linear interpolation option built into Jmol. The need for such morph animations is becoming more common as more structures are solved in different conformations. Having it built into Jmol would save a lot of time and make this kind of visualization more accessible to a much wider authorship. Thanx for this elaborate discussion, though I haven't read it in detail yet. But I do already want to mention an concept triggered by your email, and might also be mentioned in your email... Some long time ago I was discussion interpolation with someone (might have been Hens...?), and this someone suggested that a linear interpolation is fine, *especially* if done on the internal coordinates instead of the Eucledian ones. Using the internal coordinates ensures that bond length and angles are reasonable for the interpolated steps. Quite possible Egon. Our examples were about small molecules, and the animations often involved rotations of groups in the molecule. In these cases a linear interpolation of cartesian coordinates is not what you want. Maybe the four-point spline interpolation suggested by Michael would work better, but we found it simpler to do the interpolation outside Jmol (or Chime then). Starting from a MOPAC multi-structure file in internal coordinates, we use(d) a script to produce any number of intermediate structures. This method doesn't seem practical in the case of moving proteins though. Or do you think the sequence PDB -- internal -- interpolate -- PDB feasible? Hens - Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT business topics through brief surveys -- and earn cash http://www.techsay.com/default.php?page=join.phpp=sourceforgeCID=DEVDEV ___ Jmol-users mailing list Jmol-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/jmol-users
Re: [Jmol-users] Interpolation in animation?
I just noticed this thread. Since I've done dozens of interpolated animations in Chime, I'll add my 2 cents. 1. I would be pleased to see a simple linear interpolation option built into Jmol. The need for such morph animations is becoming more common as more structures are solved in different conformations. Having it built into Jmol would save a lot of time and make this kind of visualization more accessible to a much wider authorship. 2. I'm not convinced that spline smoothing or other more complicated options are worthwhile. When something more than a simple linear interpolation is needed, what is needed is usually much more complicated than smoothing. In my mind, there are three categories of conformational changes from the perspective of visualization: A. Small movements of a few residues that can be satisfactorily visualized by simply toggling back and forth between two models (empirically known, or theoretical). An example is the changes in antibody sidechain positions when antigen binds. B. Larger movements that require interpolation (morphing) in order to understand the changes in conformation. The first morph I did (recoverin +/- calcium) illustrates such a case (URL below). C. Huge movements that require that intermediate key frames be manually modeled in order to produce a useful morph. Simple linear interpolations may move domain in implausible directions, or move domains through one another. Positioning the domains in reasonable intermediate key frames, then interpolating between these key frames, gives a more useful result. An example is RNA polymerase (look for the movie and animation links at http://www.chem.umass.edu/~cmartin/ -- this is not in Chime or Jmol -- evidently they didn't publish that form?) Another dramatic example that I have long wanted to see done well is the inactivation of trypsin by a serpin. As far as I recall, it hasn't been done. For categories B and C, a linear interpolation usually has clearly bogus intermediates, such as the collapse and reexpansion of domains, the formation of cobweb clumps of bogus covalent bonds when atoms move too close to each other, or the passing of one chain through another. The latter occurs in my recoverin morph. These chemically impossible intermediates are still often compatible with a large advantage for effective comprehension of the conformational changes. Further, they serve to remind the viewer that the intermediates are not attempts to predict the true trajectory of change, but simply visual aids. When one wished to go beyond linear interpolation, the next step is to do molecular mechanics/dynamics on the interpolated intermediates to make each step chemically possible. Luckily, years of work on such routines have been provided as a free server by Werner Krebs, Mark Gerstein, and Nat Echols, the Yale Morph Server (http://www.molmovdb.org/morph/). At their Database of Macromolecular Movements (molmovdb.org) you can also find hundreds of morphs done by visitors to the site, shown in Jmol or Chime or Protein Explorer. Doing a morph with this server takes a bit of time, so I always try linear interpolation first. 3. MORPH RESOURCES (not linked above): My original Protein Morpher. This has linear and minimized interpolations of recoverin, integrin CD11b MIDAS domain, influenze hemagglutinin, calmodulin, cAMP kinase. It is available through MolviZ.Org, or go directly to http://www.umass.edu/microbio/chime/morpher/index.htm Unfortunately, it ONLY WORKS IN NETSCAPE 4.8 with CHIME. However, if you don't want to install Netscape 4.8 (instructions: http://www.umass.edu/microbio/chime/neccsoft.htm#browsers ) and Chime, you can still read: Morph Rationale and History (First? Morph): http://www.umass.edu/microbio/chime/morpher/fs_mtop.htm and Morphing Methods: http://www.umass.edu/microbio/chime/morpher/morphmtd.htm -- Protein Explorer has a control panel named NMR Models/Animation. If a morph is provided as an NMR-style multiple model PDB-format file, Protein Explorer can animate it. As you likely appreciate, displaying an animation in Chime or Jmol means you can rotate it to view the animation from different perspectives, and change the color scheme or rendering. These are in contrast to an animated GIF (true movie) which has only a single color scheme, rendering, and perspective. Protein Explorer has convenience buttons to use common color schemes and renderings for the animations. There is a script box in which you can customize the animations. Protein Explorer displays the animation in a separate browser window, which can be saved to disk. This enables the animation to be played separately, without Protein Explorer. To see many examples (as animated GIFs, and as interactive PDB file morphs), go to ProteinExplorer.Org and at the FrontDoor, click on Morphs. Or go directly to http://www.umass.edu/microbio/chime/pe_beta/pe/protexpl/morfdoc.htm Since
Re: [Jmol-users] Interpolation in animation?
On Tuesday 25 July 2006 03:58, Miguel wrote: On Jul 24, 2006, at 3:37 PM, Miguel wrote: I suggest that if you want to add this to Jmol 10 that it may be easier to add a separate 'interpolate' step that actually calculates and stores the intermediate positions of the atoms. Would you ming elaborating a bit on this? As I recall, one of the problems with 'interpolation' in Jmol v9 animations was that the locations of the atoms were not actually calculated and stored. They were. Jmol v6-v9 (maybe even earlier releases, not sure when the interpolation was hacked in) use linear interpolation, with the goal of a more smooth fitting later, which I never implemented, as the linear code was for many application already reasonable (just don't make the time steps too large before before input into Jmol). Rather, the locations of the atoms were dynamically calculated as each frame was drawn. I'm quite sure I the interpolated frames were new extra models, instead of dynamic calculation. At least that is how it was originally implemented, but maybe I missed some recoding here... At the time I thought that this was too complicated ... because it meant that the code to calculate the position in real-world Angstrom coordinates was comingled with the code to transform into screen coordinates AND, to a certain extent, the code for rendering. Additionally, the code needed to know a lot about the 'model' representation, since the interpolation took place between two models. The original implementation assumed all 'model's to have an equal number of atoms, and to be sorted equally, so that atom 1 in all frames was always the same atom. I can't remember to have fiddles with real-world/screen coordinates when doing interpolations... In order to reduce this complexity I think it would be better to break this up into pieces. I thought about something like a separate 'interpolate n' command where n was the number of steps to introduce between existing models. I was thinking that it would actually introduce new models. That way the interpolation code could focus on interpolation, and not have to worry about screen transformation issues. Right. The biggest problem that I see with this approach is that you would be inserting new models between existing models. That would require some rework of the model/frame code. So, as I am writing this it occurs to me that maybe the other approach isn't so bad. As an example of how this would work, you can take a look at the vibration code. That code dynamically applies the vector to adjust the atom position immediately before the transformation. It works well for the vibrations because the location of each atom is transformed individually. If you wanted to apply this to animations, then I think that you would want for the atom in frame n to be used from frame n - 0.5 to frame n + 0.5, where I am using a fractional frame value to represent an interpolated value. At the beginning it would be easiest to use linear interpolation between the two frames ... so the location of an atom at frame 1.25 would be the location at frame 1 + 0.25 * vector from frame 1 to frame 2. Doesn't this double the required calculations? Two instead of one, for each frame? Calculation of vector(fr1, fr2) for linear interpolation is quite cheap, but for smoother interpolations, if we really need them (see above), this might be bad... but this computation time might be neglectible with the actual time needed for redering... If you could structure it this way, I think it would be best if a the interpolation routine was written to take 4 atom positions ... to interpolate between atoms n n+1 you would also pass in frames n-1 and n+2. This would allow a smoother cubic spline interpolation that would give better results. Even if you don't implement this at the beginning, try to structure the code this way so that the parameters will be available in the future. Sound good. Some other problems that always bugged me related to interpolating animations between frames ... - you will need to interpolate atom sizes as well - you may need to interpolate atom colors as well - it isn't clear to me how other atom characteristics should be handled As Jmol v9 assumed the same atom, coloring used to be no problem... and it never dealt with different atom sizes depending on bonding or charge... not on interpolation, at least... I think that these things may also have pushed me towards the separate 'interpolate' command ... because these issues potentially introduce a lot of complexity that would be hard to handle as part of the rendering cycle. A separate interpolate command sounds good I realize that my comments may still be too hard to understand. Give some thought to the problem and please feel free to ask specific questions. Egon -- [EMAIL PROTECTED] Cologne University Bioinformatics Center (CUBIC) Blog:
[Jmol-users] Interpolation in animation?
Does 10.x support interpolation between frames in animation? I have heard that Jmol 9 supported this, but I can't seem to find information about it anywhere in 10.x. Was this feature discontinued? - Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT business topics through brief surveys -- and earn cash http://www.techsay.com/default.php?page=join.phpp=sourceforgeCID=DEVDEV ___ Jmol-users mailing list Jmol-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/jmol-users
Re: [Jmol-users] Interpolation in animation?
Eric, What does that mean, interpolation between frames? I suspect the answer is no, and I can't imagine that Jmol 9 did it. Bob eric capps wrote: Does 10.x support interpolation between frames in animation? I have heard that Jmol 9 supported this, but I can't seem to find information about it anywhere in 10.x. Was this feature discontinued? - Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT business topics through brief surveys -- and earn cash http://www.techsay.com/default.php?page=join.phpp=sourceforgeCID=DEVDEV ___ Jmol-users mailing list Jmol-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/jmol-users - Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT business topics through brief surveys -- and earn cash http://www.techsay.com/default.php?page=join.phpp=sourceforgeCID=DEVDEV ___ Jmol-users mailing list Jmol-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/jmol-users
Re: [Jmol-users] Interpolation in animation?
By interpolation between frames, I mean: an option, in the animation menu, to create a number of transition frames between two frames in an animation to make it look smoother. This is how it was described to me: I don't know how it work under the hood, but in the jmol 9 that I use, when you open the Extras-Animate panel, you can check 'interpolate between frames', and choose the number of frames that it will add between the ones you have loaded. Googling it turned up this alone: http://thomas.kuehne.cn/references/ jmol.diff. Weird? Maybe this was a branch that didn't catch on or something, but if it in fact is not implemented in Jmol 10, I could take a stab at it. On Jul 24, 2006, at 12:33 PM, Bob Hanson wrote: Eric, What does that mean, interpolation between frames? I suspect the answer is no, and I can't imagine that Jmol 9 did it. Bob - Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT business topics through brief surveys -- and earn cash http://www.techsay.com/default.php?page=join.phpp=sourceforgeCID=DEVDEV ___ Jmol-users mailing list Jmol-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/jmol-users
Re: [Jmol-users] Interpolation in animation?
By interpolation between frames, I mean: an option, in the animation menu, to create a number of transition frames between two frames in an animation to make it look smoother. This is how it was described to me: I don't know how it work under the hood, but in the jmol 9 that I use, when you open the Extras-Animate panel, you can check 'interpolate between frames', and choose the number of frames that it will add between the ones you have loaded. Yes, Jmol 9 had this feature. It did linear interpolation of atom positions between frames. Googling it turned up this alone: http://thomas.kuehne.cn/references/ jmol.diff. Weird? Maybe this was a branch that didn't catch on or something, but if it in fact is not implemented in Jmol 10, I could take a stab at it. Most of the internal architecture was redone with Jmol 10. This feature related to animation was lost. I suggest that if you want to add this to Jmol 10 that it may be easier to add a separate 'interpolate' step that actually calculates and stores the intermediate positions of the atoms. I also suggest that you may want to consider using a spline-type interpolation rather than a linear interpolation. I think that would give better interpolated results in that the interpolated positions would be a smoothed position across 4 points rather than a linear position between 2 points. Miguel - Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT business topics through brief surveys -- and earn cash http://www.techsay.com/default.php?page=join.phpp=sourceforgeCID=DEVDEV ___ Jmol-users mailing list Jmol-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/jmol-users
Re: [Jmol-users] Interpolation in animation?
On Jul 24, 2006, at 3:37 PM, Miguel wrote: I suggest that if you want to add this to Jmol 10 that it may be easier to add a separate 'interpolate' step that actually calculates and stores the intermediate positions of the atoms. Would you ming elaborating a bit on this? -Eric - Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT business topics through brief surveys -- and earn cash http://www.techsay.com/default.php?page=join.phpp=sourceforgeCID=DEVDEV ___ Jmol-users mailing list Jmol-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/jmol-users
Re: [Jmol-users] Interpolation in animation?
On Jul 24, 2006, at 3:37 PM, Miguel wrote: I suggest that if you want to add this to Jmol 10 that it may be easier to add a separate 'interpolate' step that actually calculates and stores the intermediate positions of the atoms. Would you ming elaborating a bit on this? As I recall, one of the problems with 'interpolation' in Jmol v9 animations was that the locations of the atoms were not actually calculated and stored. Rather, the locations of the atoms were dynamically calculated as each frame was drawn. At the time I thought that this was too complicated ... because it meant that the code to calculate the position in real-world Angstrom coordinates was comingled with the code to transform into screen coordinates AND, to a certain extent, the code for rendering. Additionally, the code needed to know a lot about the 'model' representation, since the interpolation took place between two models. In order to reduce this complexity I think it would be better to break this up into pieces. I thought about something like a separate 'interpolate n' command where n was the number of steps to introduce between existing models. I was thinking that it would actually introduce new models. That way the interpolation code could focus on interpolation, and not have to worry about screen transformation issues. The biggest problem that I see with this approach is that you would be inserting new models between existing models. That would require some rework of the model/frame code. So, as I am writing this it occurs to me that maybe the other approach isn't so bad. As an example of how this would work, you can take a look at the vibration code. That code dynamically applies the vector to adjust the atom position immediately before the transformation. It works well for the vibrations because the location of each atom is transformed individually. If you wanted to apply this to animations, then I think that you would want for the atom in frame n to be used from frame n - 0.5 to frame n + 0.5, where I am using a fractional frame value to represent an interpolated value. At the beginning it would be easiest to use linear interpolation between the two frames ... so the location of an atom at frame 1.25 would be the location at frame 1 + 0.25 * vector from frame 1 to frame 2. If you could structure it this way, I think it would be best if a the interpolation routine was written to take 4 atom positions ... to interpolate between atoms n n+1 you would also pass in frames n-1 and n+2. This would allow a smoother cubic spline interpolation that would give better results. Even if you don't implement this at the beginning, try to structure the code this way so that the parameters will be available in the future. Some other problems that always bugged me related to interpolating animations between frames ... - you will need to interpolate atom sizes as well - you may need to interpolate atom colors as well - it isn't clear to me how other atom characteristics should be handled I think that these things may also have pushed me towards the separate 'interpolate' command ... because these issues potentially introduce a lot of complexity that would be hard to handle as part of the rendering cycle. I realize that my comments may still be too hard to understand. Give some thought to the problem and please feel free to ask specific questions. Miguel - Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT business topics through brief surveys -- and earn cash http://www.techsay.com/default.php?page=join.phpp=sourceforgeCID=DEVDEV ___ Jmol-users mailing list Jmol-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/jmol-users
Re: [Jmol-users] Interpolation in animation?
I would be interested in hearing more about what Eric means by interpolation. An analogy to graphic data seems appropriate. A distinction I would bring up is between taking a set of dots on a graph and (a) connecting them with line segments or (b) drawing the best-fit curve through them. In the first case you get a fairly unsatisfying set of interpolations. In the second you get a smooth curve at the expense of adjusting the data. NOW, if your interest in interpolation is to just do (a) -- make a linear interpolation between each frame -- I think that might be less interesting. If, on the other hand you are interested in actually creating a time-based best fit curve through the data -- basically fitting each atom through a sequence of frames, determining a smooth curve, and then redoing the frames with interpolated frames to create the best-fit animation, THAT might be more interesting. Another possibility is to implement a moveto-like transition between frames. The idea here would be that the positions you see are interpolations of frames while the animation is running, but you can only stop at actual frames. Starting in a given frame, something like: FrameTo target number of interpolations per frame overall frames_per_second The problem there to sort out would be what to do about helixes and such. These really should be recalculated at each point. That might be asking for a lot, but maybe not. It's interesting that Jmol can calculate helixes and sheets -- it doesn't need to be told where they are. And it's very fast. I worry about the overhead, though. AH how to do this without messing up the data? How about this: a special frame that gets LOADED with atom data and it is just that animation frame that you are seeing. One frame. Perhaps with a special number, like . (When multiple files are loaded, then frames come in with numbers like 1001, 1002, 2001, 2002, so wouldn't be out of the question.) You could even define the frame number using some simple SET command. set animationFrameNumber Then perhaps one could say: frame smoothAnimation firstFrameNo lastFrameNo interpolationsPerFrame interpolationsPerSecond otherOptions The cool thing about this is that you could do all the manipulation you want -- rotations, zooming, etc. -- while the thing is playing, and it would work just fine. Basically the rendering/animation engine would work just like ever, with the data changing under it. So my proposal would be a single frame that is a clone of the atom set and is loaded on the fly. I think I've successfully separated the rendering functions from the calculation and visibility issues. There are no longer any aspects of visibility that are being determined DURING rendering - so we can define the state of the system at any time. If necessary, we could do a precalculation, but I suspect the real issue there is scalability. You can't just preload 10,000,000 atom locations, helix information, etc. But, on the other hand, there might be a way to stream this -- to save an animation in some format using the application and then stream it in and visualize on the fly. Moving the atoms is no real problem. Miguel's analogy to vibration is intriguing. One could define this all in terms of differences and the sequence of getting from frame A to frame B as, in essense, a once-through vibration vector. Interesting idea. Bob eric capps wrote: By interpolation between frames, I mean: an option, in the animation menu, to create a number of transition frames between two frames in an animation to make it look smoother. This is how it was described to me: I don't know how it work under the hood, but in the jmol 9 that I use, when you open the Extras-Animate panel, you can check 'interpolate between frames', and choose the number of frames that it will add between the ones you have loaded. Googling it turned up this alone: http://thomas.kuehne.cn/references/ jmol.diff. Weird? Maybe this was a branch that didn't catch on or something, but if it in fact is not implemented in Jmol 10, I could take a stab at it. On Jul 24, 2006, at 12:33 PM, Bob Hanson wrote: Eric, What does that mean, interpolation between frames? I suspect the answer is no, and I can't imagine that Jmol 9 did it. Bob - Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT business topics through brief surveys -- and earn cash http://www.techsay.com/default.php?page=join.phpp=sourceforgeCID=DEVDEV ___ Jmol-users mailing list Jmol-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/jmol-users - Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay
