ProFIT prediction results in CASP2

Domingues F., Floeckner H., Jaritz M. and Sippl M. Center of Applied Molecular Engineering University of Salzburg

Current methods for Protein structure prediction were tested at the
Second Meeting on the Critical Assessment of Techniques for Protein
Structure Prediction (CASP2).

Predictions were submitted for protein structures soon to be released by
experimentalists. The meeting took place last December, when the
predictions were compared to the experimental models by independent
accessors. 152 groups participated in the experiment and 947
predictions were accepted.

We participated in the Fold Recognition category, where the goal is to
identify a fold related to a target sequence in a database of known
structures. Also, the targets have no considerable sequence homology
to the known folds. We used ProFIT to produce the prediction models, and
PROSAII to evaluate them. There were 7 targets whose folds were
recognizable. We submitted predictions for 6 of these targets. We only
submitted an additional prediction for a target that turned out to be
a new type of fold, therefore not predictable by fold recognition. Of
these 6 predictions, 3 of them were considered "correct" by the
accessors, so that our model shared high structure similarity to the
experimental one. In one case (t0031) our model was considered by the
accessors to be the "best" accepted prediction. Besides these
"correct" predictions, we also predicted the correct fold type for
target t0038. The predicted model had considerable structure homology
to the target (5 out of 9 beta strands can be superimposed to 2 A). We
failed to predict two targets. Both these targets of them were large
multiple domain proteins.

Back to the list of abstracts