Building Models With Swiss-Model

Swiss-Model is an automated model server created by Manuel C. Peitsch and Nicolas Guex of the Glaxo Wellcome research centre in Geneva.

How does it work?

Essentially Swiss-Model automates the procedure described above. Full details are available at the Swiss-Model web site. Swiss-Model generates a model in 5 steps:

  1. A BLAST search finds known structures from the PDB which are similar to the target sequence
  2. From this set, parent template structures are identified which have >25% sequence identity with the target sequences and consist of at least 20 residues. In addition unrelated template regions are identified for domains which cannot be modelled in the basis of related parents.
  3. Input control files are generated from these results for the ProModII program which actually generates the models.
  4. ProModII is run to create models. To build the core, this averages the templates weighting the average by local sequence similarity. Non-conserved loops are constructed by searching the PDB for loop regions which could be accomodated on the framework. Regions which fit the framework and don't clash with other regions are sorted based on their fit to the framework and the top 5 are accepted as possible solutions. This procedure only builds the C-alpha atoms; the rest of the backbone atoms are built from a library of overlapping runs of 5 residues constructed from high-resolution structures. Sidechains are then constructed from a rotamer library.
  5. A molecular mechanics program (Gromos96) is used to energy minimise the resulting models.

How do I use Swiss-Model?

There are a number of different ways to use Swiss-Model depending on how much control you want over the modelling procedure:

Access to Swiss-Model