Skip to content

Web server manual

WHISCY

The WHISCY web server at https://wenmr.science.uu.nl/whiscy predicts the protein-protein interface of one protein chain and gives the active and passive residues to use as restraints in HADDOCK. You need an account on the portal to submit a job.

How WHISCY works

WHISCY is primarily based on conservation, but it also takes into account structural information:

  1. WHISCY finds the master sequence: the sequence of the alignment that best matches the structure.
  2. It estimates the sequence distance (amount of mutation) between the master sequence and every other sequence. This sets the amount of mutation expected at each position.
  3. For each surface residue, it compares the expected with the observed mutation. Less change than expected means conservation, and a positive WHISCY score.
  4. Interface propensities adjust the scores: phenylalanines, for example, are often in protein-protein interfaces and get a higher score; lysines rarely are and get a lower score.
  5. Surface smoothing spreads the scores over the surface of the structure: interfaces form patches, so the neighbours of an interface residue are often interface residues too.

Input

Structure

The structure of your protein as a PDB file (.pdb, at most 10 MB).

Chain

The chain to predict on. Many PDB files contain several molecules (chains); WHISCY predicts on one. The chain must be in the ATOM records of the file.

A multiple sequence alignment of your protein with homologous sequences, in one of these formats, recognised by the file extension:

Extension Format
.fasta FASTA alignment (default output of MUSCLE)
.aln CLUSTAL
.msf MSF
.phylseq PHYLIP
  • The first sequence of the alignment must be the sequence of your protein. This can be the sequence of the PDB structure, but only if it is not engineered. The structure may cover only part of that sequence.
  • A FASTA file must be aligned: at least 2 sequences, all of the same length (with gaps).
  • WHISCY is robust to a bad alignment quality, so manual correction of the alignment is not essential.

Without an alignment, WHISCY builds one from the homologs found by the ColabFold MMseqs2 server or, if it is unavailable, by EBI BLAST against UniProtKB. A job with an alignment takes a few seconds; without one, it usually takes a few minutes.

Options

  • Use interface propensities (on by default).
  • Use surface smoothing (on by default). Smoothing requires interface propensities.

Output

The results are shown online and can be downloaded for 14 days. WHISCY gives a prediction score to every surface residue that can be matched to the alignment: the higher the score, the more likely the residue is in a protein-protein interface.

Report

The results page shows the WHISCY report (report.html):

  • a 3D viewer with three views: the surface coloured by WHISCY score, from blue (lowest) to white (0) to red (highest), and, with Next, the HADDOCK residues of each selection (score cutoff, then top 10), active in red and passive in green;
  • the active and passive residues for HADDOCK, for both selections;
  • a table of the scores of all residues, which can be sorted by clicking a column header.

For an example, see the example report of the 1PPE chain E prediction.

Residues for HADDOCK

WHISCY gives two selections of active residues:

  • Score cutoff (active.txt): residues with a WHISCY score of at least 0.18. Some proteins get few or no active residues this way.
  • Top 10 (active_top10.txt): the 10 best-scored residues.

For each, the passive residues (passive.txt, passive_top10.txt) are the other surface residues with an atom within 6.5 Å of the Cα of an active residue. The files are comma-separated lists of residue numbers, ready to paste in HADDOCK.

Files

The download (whiscy.zip) contains all the files of the job, among which:

File Content
report.html, state.mvsj The report and its 3D views
<name>.pscons The WHISCY scores, one residue per line (see below)
<name>_bfactor.pdb The structure, WHISCY scores as B-factors
haddock-cutoff.pdb, haddock-top10.pdb HADDOCK residues as B-factors
active.txt, passive.txt The residues for HADDOCK (score cutoff)
active_top10.txt, passive_top10.txt The residues for HADDOCK (top 10)
whiscy.log The log of the job

Without surface smoothing the scores file is <name>.acons, and without either option <name>.cons.

To view report.html from the download, serve its directory with python -m http.server -d <directory> and open http://localhost:8000/report.html: browsers do not let a page opened from disk load the files of the 3D viewer.

If a job fails

The results page shows the error output of the job, which often says why it failed (for example an alignment whose first sequence does not match the structure). For help, ask at https://ask.bioexcel.eu.

Interpreting the predictions

You should not trust WHISCY predictions blindly, but always try to relate them to information that you already have.

  • WHISCY works very well for proteins with a single conserved protein binding site.
  • It should not be used for non-conserved interactions, such as antibody-antigen.
  • If you know that your protein has multiple binding sites, verify whether a predicted site could be the one you are interested in.

Running WHISCY locally

WHISCY is also available as a stand-alone program: see Installation and Usage.

References

To cite WHISCY, please refer to: