The Robetta server

2.1. Preserving existing functionality 548

An unparalleled amount of sequence data is being made available from large-scale genome sequencing efforts. The data provide a shortcut to the determination of the function of a gene of interest, as long as there is an existing sequenced gene with similar sequence and of known function. This has spurred structural genomic

design of macromolecular

Analyzing and engineering cellular signaling processes requires accurate estimation of cellular subprocesses such as protein-folding. We apply parametric and nonparametric classification to the problem of assessing three-dimensional protein domain structure predictions generated by the Rosetta ab initio structure prediction method. The assessment is based on whether the predicted structure is similar enough to a known protein structure to be classified as being in the same protein superfamily. We develop appropriate features and apply Gaussian mixture models, K-nearest-neighbors, and the recently developed linear interpolation with maximum entropy method (LIME). The proposed learning methods outperform a previous quality assessment method based on generalized linear models. Results show that the proposed methods reject the vast majority of poor structural predictions while identifying a useful number of good predictions. 1.

Saccharomyces cerevisiae is one of the best-studied model organisms, yet the three-dimensional structure and molecular function of many yeast proteins remain unknown. Yeast proteins were parsed into 14,934 domains, and those lacking sequence similarity to proteins of known structure were folded using the Rosetta de novo structure prediction method on the World Community Grid. This structural data was integrated with process, component, and function annotations from the Saccharomyces Genome Database to assign yeast protein domains to SCOP superfamilies using a simple Bayesian approach. We have predicted the structure of 3,338 putative domains and assigned SCOP superfamily annotations to 581 of them. We have also assigned structural annotations to 7,094 predicted domains based on fold recognition and homology modeling methods. The domain predictions and structural information are available in an online database at

ABSTRACT A new multibody, whole-residue potential for protein tertiary structure is described. The potential is based on the local environment surrounding each main-chain � carbon (CA), defined as the set of all residues whose CA coordinates lie within a spherical volume of set radius in 3-dimensional (3D) space surrounding that position. It is shown that the relative positions of the CAs in these local environments belong to a set of preferred templates. The templates are derived by cluster analysis of the presently available database of over 3000 protein chains (750,000 residues) having not more than 30 % sequence similarity. For each template is derived also a set of residue propensities for each topological position in the template. Using lookup tables of these derived templates, it is then possible to calculate an energy for any conformation of a given protein sequence. The application of the potential to ab initio protein tertiary structure prediction is evaluated by performing Monte Carlo simulated annealing on test protein sequences. Proteins 2004;00:000–000. © 2004 Wiley-Liss, Inc. Key words: packing motifs; structural motifs; multibody potentials; statistical potentials; simulated annealing; structure prediction

The most significant impediment for protein structure prediction is the inadequacy of conformation space search methods. Conformation space is too large and the energy landscape too rugged for existing search methods to consistently find near-optimal minima. To alleviate this problem, we present model-based search, a novel conformation space search method. Model-based search uses highly accurate information obtained during search to build an approximate, partial model of the energy landscape. Model-based search aggregates information in the model as it progresses, and in turn uses this information to guide exploration towards regions most likely to contain a near-optimal minimum. We validate our method by predicting the structure of 32 proteins, ranging in length from 49 to 213 amino acids. Our results demonstrate that model-based search is more effective at finding lowenergy conformations in high-dimensional conformation spaces than existing search methods. The lower-energy conformations found by our method also correspond to higher-accuracy structure predictions. 1

Research article Protein-segment universe exhibiting transitions at intermediate segment length in conformational subspaces

Research article Universal partitioning of the hierarchical fold network of 50-residue segments in proteins