The LabelHash Server

Help

This server runs the LabelHash algorithm as described in our CSB 2008 paper. Given a point-based motif, it will compute all matches in either the full PDB or a non-redundant version of the PDB (with less than 95% sequence identity), both from November 11, 2008. The points in this case are the C-alpha coordinates of a number of residues. It is possible to specify alternate residue labels for each motif point. Computing matches in the full PDB is about 5 times slower than computing matches in the non-redundant PDB.

On the main page, fill out the form to specify a motif. If you click submit, your job will be submitted to a queue. You will receive an email message once your job starts. Once your job is finished, you will receive another email message with a file of matches attached. You can visualize the file with matches using our ViewMatch plugin for Chimera. You can also look at them as plain text using the command:
xmllint -format mymatches.xml | less
Xmllint is part of the libxml2 package and is often already installed on UNIX-like operating systems such as Linux and OS X. If you would like to run LabelHash on your own machines, download the command line version of LabelHash.

Sample Output

Below are some example motifs that were used to test the LabelHash algorithm. For each motif we have pre-computed the 100 best matches in the entire PDB from November 11, 2008. For each motif we list the EC class it represents, the PDB ID of the protein that was used to get the 3D coordinates of the motif, and the motif residues. The residues are specified by residue ID in the corresponding PDB file. For each residue the allowed residue types are denoted in superscript by their one-letter residue abbreviations. So if only one residue type is listed, no substitutions are allowed. The last column has a link to the corresponding file of matches. Right-click on the link to save these match files on your computer. These files can be opened with the ViewMatch plugin for Chimera. You can also look at them as plain text using your browser or an editor.

An XML match file is structured as follows:

• First, it lists the number of matches that were found and the time it took.
• Next, it lists the motif, the matching parameters, and the list of LabelHash tables that the motif was matched against.
• Finally, the bulk of the file consists of a list of matches. For each match the following information is listed:
  • The reference sets in the motif and the match. These are the partial matches that formed the starting point for matching augmentation. See the CSB2008 paper for details.
  • The alignment between the motif and the match. This is specified by a vector of 7 numbers. The first 4 numbers specify the rotation as a unit quaternion. The last 3 numbers specify the translation.
  • The score indicates the number of matched residues. The web server only returns complete matches, so this number is not very informative. With the command line version, however, one can compute partial matches.
  • The rmsd is the LRMSD between the motif and the match.
  • The depth is the average distance between the matched residues and the molecular surface. This is not currently used during matching, but it is just returned as an extra bit of information.
  • The p-value indicates the statistical significance of a match.
  • The correspondence shows the mapping between the motif residues and residues in the match. The residue numbers are based on a renumbering of the residues in the corresponding chains starting from 0.