Morph Server FAQ
From Info
Frequently asked questions regarding the morph server
Q. How realistic are these movies?
A. First, a disclaimer: these movies in no way represent what actually occurs during conformational change, although there are many cases where the real intermediates are probably quite similar. No time-scale/kinetic/dynamic information is known, and there may be additional states that have not been solved. The Morph Server is primarily a tool for informatics and scientific illustration, not a replacement for more crystallography or detailed (and time-consuming) MD simulation.
That said, the intermediate structures are usually fairly realistic, enough so to be indistinguishable in a molecular graphics program if not to PROCHECK. Energy minimization is kept to a minimum, but helps remove the most blatant distortions. The exceptions are a growing category of structures that undergo partial refolding or other massive transformations, such as the serpins or T7 polymerase, where a relatively linear pathway cannot be determined.
Q. On a very basic level, how are morphs generated?
A. The details of the server's functionality can be found distributed throughout several of MolMovDB's key published works, as well as in references describing CHARMM. A comprehensive description of how morphs are generated is difficult to provide here, but a very basic description is provided below.
Morphs are generated by a rapid adiabatic mapping procedure (simple interpolation followed by energy minimization of each intermediate), and represent a compromise between chemical reality and computational efficiency. The "morph engine” implements the CHARMM potential (originally through X-PLOR) to adiabatically interpolate between two known conformations. This constitutes an improvement to straight Cartesian interpolation, which may result in physically unrealistic intermediates (e.g. clashing atoms).
In other words, morphing is a compromise between straight Cartesian interpolation (wherein the x, y, and z, coordinate differences between each atom of 2 conformations are split into even intervals in generating intermediary states) and molecular mechanics. Each of these intermediary states (as initially generated by straight Cartesian interpolation) is energy-minimized using CHARMM. Thus, the CHARMM molecular mechanics force field, as well as the concept of energy minimization in general, are central to morph generation.
Q. I received an error message from the morph server. What happened?
A. There are several reasons why the server may fail. We've tried to make it obvious from the error message what exactly is wrong, but this isn't always very helpful, and the server isn't very well set up for debugging submissions. A few of the most common problems:
- Conflicting SEQRES records. This is by far the most comment cause of failure! The server uses SEQRES data to determine the actual protein sequence. However, this frequently conflicts with the sequence intuited from ATOM records (including official PDB files from www.rcsb.org). If the server cannot automatically resolve these conflicts, it will fail. The easiest workaround to this is to submit files with SEQRES records deleted; if you supplied a PDB ID rather than a file you will need to download the proper file from the PDB, then modify and upload it. However, this may sometimes lead to other distortions, depending on the sequence numbering.
- Misformatted files. Although we're not sure of the exact details, files that do not have the standard Unix newline format usually are not parsed correctly. This is only a problem with user-submitted files, and can be fixed fairly easily with utilities such as dos2unix. (We'll probably modify the server to handle this too.)
- Incorrect parameters. If you enter a chain ID that is not present in the submitted files, this will result in an error when the server tries to fit the files.
- Server error. Sometimes we find (gasp!) bugs in the server. In these cases, we will of course attempt to fix the problem ourselves. Since the administrator receives notice of all morph failures, he is usually able to check and find out why the server quit.
Note that the error messages differ greatly in the development version.
Q. Why does the final sequence differ from the protein I input?
A. The actual morphing is performed by X-PLOR. For this to work, the pair of input structures must i) be identical in sequence; ii) start numbering residues at 1 and not have any gaps; iii) must be a single chain (not counting heteroatoms). If original files do not meet these requirements, the server will attempt to homogenize them automatically. Usually this results in only minor changes (if any).
Q. I have a really complex structure / top-secret coordinates / special movie needs - can you help me?
A. We strongly discourage private submissions because they go against the spirit of the database, which is not only intended to provide free morphing to individual users, but also to be a browseable and searchable repository of morphs which are useful to others. We understand, however, that some morphs may reveal confidential information and so beyond moral suasion nothing prevents you from using the "Private" check box on our single- and multi-chain morph submission forms. This sets a flag in our database that tells our movie gallery page not to display the morph. Also, the search tool on our front page will not return it. The only way a member of the public could possibly find your morph is if they were able to intercept the email you got from our server with the link to it. We consider the probability that this will happen to be very low. Generally we cannot handle specific requests for further security unless as part of an official collaboration. We can, however, give you pointers on generating more sophisticated or customized morphs. In many cases we prefer that you obtain our software and create a morph yourself. This applies to many of the questions below as well.
Q. Why do my final structures look like they've been through a blender?
A. This is a problem with automatic sequence determination. Though SEQRES records may cause errors due to gaps or misnumbering in the ATOM records, the ATOM records alone are sometimes not the most trustworthy indicator of sequence. The server may process gaps or numbering issues incorrectly, which causes the structure to be distorted during the homgenization step. This is easily fixed, but there are several different approaches that may be necessary; email the administrators for help. So far, the beta server does not seem to have this problem, although it has other limitations.
Q. The protein appears to expand and contract in the movie.
A. We're not sure what causes this. We think it has something to do with the absence of explicit hydrogen atoms and the effect of this on energy minimization, but it pops up inconsistently.
Q. What about heteroatoms?
A. Heteroatom processing is extremely difficult, usually because they're either missing or differently numbered in one of the files. Automatically collecting the necessary parameters for X-PLOR/CNS has proven thorny as well. We have enabled heteroatoms in the automatic web form, but the server may make incorrect guesses about their relative locations. (Okay, currently [April 2003] this has been disabled entirely. Contact us privately if you really need to include heteroatoms.)
Q. What about nucleic acids?
A. There's no reason why the server can't support DNA or RNA structures, except that it requires a fair amount of time and images have to be done somewhat differently due to limited support by graphics software. We now have limited support for RNA and DNA through the multichain version of the server, and the beta version will incorporate this soon. Constraints on this method are discussed elsewhere. Early results have been excellent, except that the graphics software doesn't handle nucleic acid ribbons at all.
Q. Can I get a morph without reoriented structures?
A. Only with the beta server, which makes structural alignment optional. For files directly from the PDB, such alignment is often essential, e.g. if the structures are from different crystal forms. If you already have structures in the desired orientation, however, you may submit these to the beta server with fitting turned off.
Q. My PDB file has multiple chains. How do I get a morph of the entire structure?
A. Use the beta version. See the help document linked from the input form for details.
Q. My structure has gaps, but the morph server eliminates these. Is there any way to avoid this?
A. This can also be fixed with the beta version. Note that all gaps in the first structure will be preserved, but if the second structure is not 100% identical some gaps may be added or deleted from it.
Q. I downloaded the collection of PDB files for a morph, but I can't figure out how to expand the archive.
A. The downloaded file is in GZIPed UNIX tar format, i.e. its extension should be .tar.gz (for the NMR format PDB file, it's only GZIPed). Unfortunately, many (most?) web browsers will not name the downloaded file appropriately. Rename it with the correct extension and open it with Winzip or gzip/tar. There is probably a simple fix for this on the server but we haven't gotten around to it.
Q. Do you distribute your software?
A. Absolutely! Well, parts of it, anyway. The Morph Server as a whole isn't really a distributable project, although if you're really interested you should contact Mark Gerstein. Most people are interested in the script which actually generates the interpolation; visit the download page for instructions and relevant files. We also have a great deal of software for analyzing macromolecular geometry that is freely available.
Q. Are there any other programs like this available?
A. Several, with varying features and purposes. However, it is also possible to construct something resembling the morph server using a suitably flexible dynamics package that allows scripting, such as CNS/XPLOR (which we use) or MMTK. You may want to have a look at our Related Resources page.
Q. After submitting jobs to the morph server, I am presented with a screen containing my morph ID, and informing me that I will be emailed once processing is complete. However, I never received this follow-up email with the link to my generated morph. What may be the problem?
A. Yale University blocks servers from directly sending out emails to non-Yale email servers. MolMovDB is in the exception list, but every year the university does a cleanup, and sometimes takes out that exception. If this is the case, please contact us if you have not received an email notice that your job has been complete, and we will correct the problem immediately.
- RigiMol. Written by PyMOL author Warren DeLano, and apparently accessible from within PyMOL. Not yet released, and we don't know how it works, but the sample movies look very good.
- LSQMAN. Structural alignment program by Gerard Kleywegt, also does simple morphing using Cartesian or "internal" coordinates.
- Indie server at the NIH. Similar to the Morph Server, but uses linear interpolation and requires the submitted files to have the same sequence.
- DynDom. Does not morph structures, but rather performs detailed structural comparisons similar to the statistics from our server. This is also distributed as part of CCP4, and has a database of motions as well.
The use of targeted molecular dynamics has also grown in recent years; such simulations add a pseudo-energy term for the RMSD relative to the "destination" structure. This allows simulation of motions that normally occur on a large time scale; although kinetic information is lost, the molecular interactions can be studied (e.g. the effect of activation loop conformation on the global conformation of Src kinase). An alternate method is steered molecular dynamics, wherein a force is applied to a selection of atoms in the direction of motion. These simulations are implemented in CHARMm and NAMD, resepctively.
-- Nat Echols (02-06-03)
