We have recently developed and validated a method for
de-novo protein structure prediction by free-energy refinement.
In this approach we generate a large library of possible conformations using a heurisitic method for protein structure prediction (presently ROSETTA). These decoys are then relaxed in our all-atom protein forcefield and ranked energetically. A small subset of the best decoys is clustered and the largest cluster is used to predict the structure.
In a recent test this protocol was successful in over 90% of the cases for monomeric single-chain proteins with less than 85 amino acids. The median RMS deviation between model and experimental conformation was 2.4 A. The figure below illustrates the overlay of several proteins that were correctly predicted.
We now seek candidates for blind prediction: Ideally the proteins should be:
- have less than 100 amino acids
- no disulfide bridges
- an do not require large ligands for stabiliuzation
To submit a sequence for prediction: mail the sequence to Wolfgang Wenzel
At the present time our protocols are not fully automated. Depending on the availability of computational resources and manpower, we will process the request submitted. If all goes well, we will mail back a structure a synopsis of the prediction process. Please mail us, in case you have questions.