Finding the needle in a haystack: Educing native folds from ambiguous ab initio protein structure predictions

Current ab initio structure-prediction methods are sometimes able to genera te families of folds, one of which is native, but are unable to single out the native one due to imperfections in the folding potentials and an inabil ity to conduct thorough explorations of the conformational space. To addres s this issue, here we describe a method for the detection of statistically significant folds from a pool of predicted structures. Our approach consist s of clustering and averaging the structures into representative fold famil ies. Using a metric derived from the root-mean-square distance (RMSD) that is less sensitive to protein size, we determine whether the simulated struc tures are clustered in relation to a group of random structures. The cluste ring method searches for cluster centers and iteratively calculates the clu sters and their respective centroids. The centroid interresidue distances a re adjusted by minimizing a potential constructed from the corresponding av erage distances of the cluster structures. Application of this method to se lected proteins shows that it can detect the best fold family that is close st to native, along with several other misfolded families. We also describe a method to obtain substructures. This is useful when the folding simulati on fails to give a total topology prediction but produces common subelement s among the structures. We have created a web server that clusters user sub mitted structures, which can be found at http://bioinformatics.danforthcent er.org/services/scar. (C) 2001 John Wiley & Sons, Inc.