The Pcons / Pmodeller meta server is based on ideas and methods we have developed in our lab the last five years. These, include methods that combines machine learning approaches with consensus analysis to select the best possible model from a set of plausble protein structures. A schematic overview of the method is shown below:
In addition, the server also automatically splits up a sequence into multiple domains if a significant hit does not span the whole target sequence. The criteria for domain split-up is: 1) a significant RPS-BLAST hit to the PDB and 2) the unaligned sequence should be longer than 100 amino acids.

Sent any a comments to Björn Wallner (bjorn@sbc.su.se)