LSD-CAFASP: Loop and SiDechains CAFASP


The goal of LSD-CAFASP is to evaluate the accuracy of loops and side-chains in protein structure prediction. An additional goal is to assess the capabilities of the tools to generate models whose structures are closer to the corresponding native structure than to the templates used.


A number of prediction targets will be selected from PDB-CAFASP. PDB-CAFASP gathers structure predictions from automatic fold-recognition servers using as targets those pre-release PDB entries, with no significant sequence similarity to proteins already in the PDB.

Those PDB-CAFASP targets for which a highly reliable structure prediction is observed (e.g. many of the fully-automated servers report similar results with above-threshold scores) will be selected for LSD-CAFASP. As has been demonstrated in CAFASP3, the highly reliable FR alignments provided by the servers are relatively accurate, and thus, can be used the basis for loop and side-chain modeling.

In LSD-CAFASP the alignments and the templates will be supplied so that all participants start their modeling using the exact same data.

After the native structures of each target is obtained, we will evaluate how accurate the selected FR alignment was. If the selected FR alignment contains regions which significantly differ from the native structure, so that a meaningful evaluation of the LSD-CAFASP predictions can not be carried out, then these regions/loops will be excluded form the evaluation.

Target LSD-CAFASP submission deadline (IDT) Sequence-Template Alignment From Server Predicted backbone model
1MOI 9-Mar-2003 fugue3.0 ali , pdb-model
1MQW 17-Mar-2003 inbgu.0 ali , pdb-model
1MRZ 19-Mar-2003 3dpsm.1 ali, pdb-model
1MZK 08-Apr-2003 orfeus.0
1N8N 21-May-2003 inbgu.0
1NE2 10-Jun-2003 fugue3.2
1J2G 05-Jul-2003 inbgu.0 ali , pdb-model
1NTE 29-Jul-2003 ffa3.0

Target LSD-CAFASP submission deadline (IDT) Sequence-Template Alignment From Server Accuracy of the selected pdb-model (MaxSub) MaxSub superposition of selected pdb-model and native
1MWM Expired fugue3.0 ali , pdb-model (NOTE: partial match) 12.7 A RMSD; MaxSub score: 2.8 (subset: 106/312 at 1.7 A RMSD)
1M33 Expired fugue3.5 ali , pdb-model 2.9 A RMSD; MaxSub score: 6.2 (subset: 197/266 at 1.7 A RMSD)
1MKE Expired ffas3.1 ali , pdb-model 3.3 A RMSD; MaxSub score: 5.2 (subset: 94/144 at 2.8 A RMSD)
1KYT Expired fugue3.2 ali , pdb-model (NOTE: partial match) 19.6 A RMSD; MaxSub score: 2.6 (subset: 67/210 at 1.5 A RMSD)


The targets will be available for prediction until the specified deadline or until their 3D structure are released. Targets will expire also if the structure of a homologous protein is released. Predictions should be submitted to in PDB format only and will be accepted anytime before the expiration time.

Upon release of the experimental structure, the predictions will be compared and assessed.

To ensure that the predictions were made using the sequence-template alignment provided by LSD-CAFASP, each prediction will be checked to see if at least 90% of their C-alphas of the core are identical to the those of structure implied by the provided alignment. Predictions that will differ by more than 10% will be disqualified from the main LSD-CAFASP evaluation, and will be evaluated in a special "different source" category.


Daniel Fischer

Leszek Rychlewski


The following is a partial list of groups who have expressed interest in participating:

  • Silvio Tosatto
  • Barry Honig
  • Jeff Skolnick
  • Roland Dunbrack


Anyone wanting to participate in this experiment should register by sending an e-mail to Daniel Fischer