Online Structure Alignment Resources (April 2005; later additions as noted)

This list is not meant to be comprehensive, but includes a reasonably broad sample of sites I think are handy and convenient to use.

The test cases involve structures with low pairwise sequence identity:

• Pairwise - 2mnr (mandelate racemase) 4enl (enolase)
• Multiple - 2mnr 4enl 1nu5 (MLE II)

Note that the MatchMaker tool in Chimera performs a similar function: it constructs a sequence alignment and superimposes the structures accordingly. As of version 1.2129 (June 2005), secondary structure can be used to help construct the alignment, which allows pairs with lower percent identity to be superimposed correctly (see results from using MatchMaker).

• FAST - pairwise; paper compares the method to DaliLite, CE, and K2.

  Reference: Zhu J, Weng Z. FAST: a novel protein structure alignment algorithm. Proteins. 2005 Feb 15;58(3):618-27.

  Availability: server; executables for Linux, IBM, Sun, Alpha, Mac OS X can be downloaded

  Performance: ~5 seconds for pairwise. Shows a sequence alignment in the web page; superimposed coordinates are available within the downloadable RasMol script (puts both into chains into one model and makes them A and B; Chimera complains about the bad PDB records but will display the structures nonetheless).

• MultiProt - truly multiple rather than assembly of pairwise alignments; can handle circular permutation. Method seems elaborate (opinion from reading the paper) but works well.

  Reference: Shatsky M, Nussinov R, Wolfson HJ. A method for simultaneous alignment of multiple protein structures. Proteins. 2004 Jul 1;56(1):143-56.

  Availability: server; can be downloaded for Linux

  Performance: ~28 seconds for 3-way using sequence order. Reports several alignments: five of all three, five with various pairings. I took the 3-way alignment with the most residues aligned. Coordinates are combined in a single file but come out as separate submodels in Chimera (#0.1, #0.2, #0.3).

• SuperPose - multiple; authors claim the server is especially user-friendly.

  Reference: Maiti R, Van Domselaar GH, Zhang H, Wishart DS. SuperPose: a simple server for sophisticated structural superposition. Nucleic Acids Res. 2004 Jul 1;32(Web Server issue):W590-4.

  Availability: server (uses a number of other programs including VADAR and ClustalW)

  Performance: ~25 seconds for pairwise (not including waiting for the display applet). Results page allows download of superposition in PDB format, sequence alignment, and RMSD information. Coordinates are combined in a single file but come out as separate submodels in Chimera (#0.1, #0.2, #0.3). Only two files can be uploaded, so according to the instructions you have to combine input files to superimpose more than two structures. I had trouble getting this to work. The support person said the structures must have chain IDs; however, adding chain IDs to my input did not work. Only taking the advice to use the server to superimpose two and then using that output (which, incidentally, did NOT include chain IDs) as one of the inputs for another round of superposition worked.

• SALIGN (added September 2005) - multiple

  Reference: Marti-Renom MA, Madhusudhan MS, Sali A. Alignment of protein sequences by their profiles. Protein Sci. 2004 Apr;13(4):1071-87.

  Availability: server

  Performance: ~15 seconds for 3-way. The resulting web page shows the superposition in a Jmol window and a sequence alignment. The sequence alignment can be downloaded in PIR, FASTA, or PAP format, and a PDB file of the superimposed coordinates can be downloaded. Coordinates are combined in a single file but come out as separate submodels in Chimera (#0.1, #0.2, #0.3). These were not in the order of specification in the input file, but the submodels were also given chain IDs (A, B, C) that did reflect that order. Rather than using that PDB file, I find it easier to open the alignment in Chimera, autoload the structures from the PDB (since the sequences are named with the PDB IDs), and match the structures using the sequences, because this retains the original chain IDs plus ligands, ions, etc.

• K2, K2SA - pairwise; the program (but not the server?) can handle circular permutation.

  Reference: Szustakowski JD and Weng Z. Protein structure alignment using a genetic algorithm. Proteins, 38(4):428-440, Mar 2000.

  Availability: server

  Performance: ~30 seconds using sequential constraints. Can download coordinates for second protein (transformed to match first protein) and a log file.

• CE (pairwise), CE-MC (multiple, building upon pairwise CE alignments)

  CE	CE-MC

  References: Shindyalov IN, Bourne PE. Protein structure alignment by incremental combinatorial extension (CE) of the optimal path. Protein Eng. 1998 Sep;11(9):739-47.
  
  Guda C, Lu S, Scheeff ED, Bourne PE, Shindyalov IN. CE-MC: a multiple protein structure alignment server. Nucleic Acids Res. 2004 Jul 1;32(Web Server issue):W100-3.

  Availability: CE server ("software" available but I can't tell what is in the tar file without downloading it), CE-MC server (binaries and source code available for Linux and Solaris)

  Performance: ~55 seconds for pairwise (CE server). ~4 minutes for multiple alignment (CE-MC); e-mail address required for this option. Results include structure-based sequence alignment and single downloadable PDB in which the structures are given different chain IDs (Chimera complains about duplicate atom serial numbers but displays the structures nonetheless). In the CE-MC results, the residue numbers are shifted over one column from where they should be, so the file has to be edited before viewing. The support person said the problem would be fixed in the next update.

• DaliLite - pairwise (the Dali server instead compares a single input structure against a representative set from the PDB, like VAST Search)

  Reference: Holm L, Park J. DaliLite workbench for protein structure comparison. Bioinformatics. 2000 Jun;16(6):566-7.

  Availability: server; download (Unix) and secondary server available from another site

  Performance: ~1 minute; reports several alignments and their stats; for each, coordinates can be downloaded for the second protein. The figure shows the superposition with the most residues aligned.

• Protein3Dfit (added July 2005) - pairwise

  Reference: Lessel U, Schomburg D. Similarities between protein 3-D structures. Protein Engineering. 1994 Oct;7(10):1175-87.

  Availability: server

  Performance: ~25 seconds for pairwise (default settings; various parameters can be adjusted by the user). The resulting web page shows the superposition in a Jmol window plus a sequence alignment and RMSD table. Links include many different ways (convenient!) to download the results. One way to save the superposition is by downloading the transformed coordinates of the second structure, and the sequence alignment can be saved in PIR format.

• later additions (not evaluated)
  • OPAAS (Optimal, Permuted and other Alternative Alignments of protein Structures)
  • 3D-SS (choice of method: STAMP or ProFit)
  • ARTS (for nucleic acid structures)
  • Mammoth (output coordinates are CA-only, but the sequence alignment can be downloaded in various formats)
  • FATCAT (Flexible structure AlignmenT by Chaining Aligned fragment pairs allowing Twists; i.e., one of the structures may be split into pieces)
  • POSA (flexible; multiple; requires e-mail address;)
  • GANGSTA (seq-order-independent; requires e-mail address)
  • SCMBB pairwise protein superposition server
  • CPalign (seq-order-independent; requires e-mail address)
  • TopMatch (successor of ProSup)
  • PDBj superposition servers (RASH and GASH) - seek to maximize NER (number of equivalent residues) at 4 Å cutoff
  
  
• servers that don't provide output PDB coordinates (as far as I can tell)
  • TOPOFIT (seq-order-independent)
  • CURVE

Test structures: 1qq5 A chain (HAD) 3chy (CheY)

These are in different superfamilies according to SCOP, but it has been noted that the structures appear to be related by circular permutation (see paper by Ridder and Dijkstra). The active site residues are in about the same positions in 3D, although in different N->C orders. Selected active site residues (those in a column are aligned in 3D):

   -----------------------------------------------------------
                loop 1     loop 2     loop 3     loop4       
   -----------------------------------------------------------
   2-HAD X. autrophicus                                  *
                 D8         S114       K147      171SS---D176
   epoxide hydrolase M. musculus, H. sapiens        *
                 D9         T123       K160      184DD---N189
   CheY E. coli, S. typhimurium                     *
                D57         T87        K109       12DD13
   -----------------------------------------------------------

superimposed manually	MultiProt (~10-15 seconds) (the choice with the most residues aligned and as a tie-breaker, the lowest RMSD)	K2SA (~15-20 seconds) (wrong, see log file)

The Multiprot results correctly place the corresponding active site residues together; the K2SA results are shifted over by one strand in the central sheet. NB: The K2SA server appears to be imposing sequential constraints even when I set the option to NOT do so... which would explain the failure on this test case. I reported the problem and plan to try again if it is fixed.

• 3DCoffee - multiple; primarily intended to generate sequence alignments but can use 3D structures to do so; can also combine multiple sequence alignments

  Multalign Viewer matching defaults	with iteration (pruning)

  Reference: Poirot O, Suhre K, Abergel C, O'Toole E, Notredame C. 3DCoffee: a web server for mixing Sequences and Structures into multiple sequence alignments. Nucleic Acids Research 2004 Jul 1;32(Web Server issue):W37-40.

  Availability: TCoffee server (includes 3DCoffee option)

  Performance: The three PDBs were submitted using the Advanced 3DCoffee page. Results were returned in ~4 1/2 minutes and include an alignment (clustal format, *.aln) and dendogram information (*.dnd). The superpositions were generated in Chimera by using the sequence alignment with and without pruning of CA pairs more than 2.0 A apart.

• S4 database (Structure-Based Sequence Aligments of SCOP Superfamilies) - 40% ID filtered subset (ASTRAL40). A drawback is that there may not be sequences for (or sufficiently similar to) the structures you want to superimpose. Also, I had problems with browsing the database, so it was easier to just download the whole thing (available as a tar file). Alignments are in clustal format (*.aln) which can be used in Chimera. Sequences are named with ASTRAL domain IDs, so you can auto-load all domains (if not too many) or just open a subset in the standard way. Example: S4 "Enolase C-terminal domain-like" superfamily page

  Reference: Casbon J, Saqi MA. S4: structure-based sequence alignments of SCOP superfamilies. Nucleic Acids Res. 2005 Jan 1;33(Database issue):D219-22.

• HOMSTRAD database (Homologous Structure Alignment Database) - proteins grouped by family (narrower groupings, likely also alignable with sequence methods including MatchMaker in Chimera). The advantage of using the database alignments even if the sequences are easily alignable is that they can contain many sequences (avoids a step of combining pairwise alignments). The database also includes superimposed coordinates, but in the annoying RasMol way (a single file with different chain IDs but many duplicate atom serial numbers). I prefer to download the PIR-format alignment and use it in Chimera. Sequences are named with PDB IDs, so you can auto-load all structures (if not too many) or just open a subset in the standard way. Example: HOMSTRAD enolase family page

  References: Mizuguchi K, Deane CM, Blundell TL, Overington JP. (1998) HOMSTRAD: a database of protein structure alignments for homologous families. Protein Science 7:2469-2471.
  
  Stebbings LA, Mizuguchi K. HOMSTRAD: recent developments of the Homologous Protein Structure Alignment Database. Nucleic Acids Res. 2004 Jan 1;32(Database issue):D203-7.

• DBAli - contains an all-against-all comparison of protein structures (September 2005: contains 888,695,103 pairwise structural alignments and family-based multiple structure alignments for 23,987 non-redundant chains). Multiple superpositions are presented just as described for SALIGN above. Pairwise superpositions are presented similarly, with sequence alignments downloadable in PIR format and superposition coordinates downloadable in PDB format (the two chains are given different IDs but combined into a single model, i.e., are not separated by END records).

  Reference: Marti-Renom, M.A. Ilyin, V.A. Sali, A. DBAli: A database of protein structure alignments. Bioinformatics 17:746-7 (2001).

• later additions (not evaluated)
  • Dictionary of Homologous Superfamilies (Multiple alignments of S35Reps available as (a) coordinates embedded in Rasmol script (b) FASTA sequence)
  • Database of Multiple Alignments for Protein Structures (DMAPS) (Multiple alignments for SCOP, CATH, ENZYME, and CE-based families generated by CE-MC superposition can be downloaded in sequence and structure form)
  • SISYPHUS structural alignments for proteins with non-trivial relationships (FASTA alignments)
  • PROMALS3D (web server takes input FASTA sequences and/or PDB structures, generates multiple sequence alignments in a few formats)