EVA: continuous automatic evaluation of protein structure prediction servers.

UNLABELLED: Evaluation of protein structure prediction methods is difficult and time-consuming. Here, we describe EVA, a web server for assessing protein structure prediction methods, in an automated, continuous and large-scale fashion. Currently, EVA evaluates the performance of a variety of prediction methods available through the internet. Every week, the sequences of the latest experimentally determined protein structures are sent to prediction servers, results are collected, performance is evaluated, and a summary is published on the web. EVA has so far collected data for more than 3000 protein chains. These results may provide valuable insight to both developers and users of prediction methods. AVAILABILITY: http://cubic.bioc.columbia.edu/eva. CONTACT: eva@cubic.bioc.columbia.edu

Protein structure prediction using a combination of sequence-based alignment, constrained energy minimization, and structural alignment.

We present a novel approach to protein structure prediction in which fold recognition techniques are combined with ab initio folding methods. Based on the predicted secondary structure, one of two different protocols is followed. For mostly alpha proteins, global optimization and sampling of a statistical energy function is used to generate many low-energy structures; these structures are then screened against a fold library. Any structural matches are then selected for further refinement. For proteins predicted to have significant beta-content, sequence and secondary structure-based alignment is used to identify candidate templates; spatial constraints are then extracted from these templates and used, along with the statistical energy function, in the global sampling and optimization program. Successes and failures of both protocols are discussed.

EVA: large-scale analysis of secondary structure prediction.

EVA is a web-based server that evaluates automatic structure prediction servers continuously and objectively. Since June 2000, EVA collected more than 20,000 secondary structure predictions. The EVA sets sufficed to conclude that the field of secondary structure prediction has advanced again. Accuracy increased substantially in the 1990s through using evolutionary information taken from the divergence of proteins in the same structural family. Recently, the evolutionary information resulting from improved searches and larger databases has again boosted prediction accuracy by more than 4% to its current height around 76% of all residues predicted correctly in one of the three states: helix, strand, or other. The best current methods solved most of the problems raised at earlier CASP meetings: All good methods now get segments right and perform well on strands. Is the recent increase in accuracy significant enough to make predictions even more useful? We believe the answer is affirmative. What is the limit of prediction accuracy? We shall see. All data are available through the EVA web site at [cubic.bioc.columbia.edu/eva/]. The raw data for the results presented are available at [eva]/sec/bup_common/2001_02_22/.

Prediction of protein tertiary structure to low resolution: performance for a large and structurally diverse test set.

We report the tertiary structure predictions for 95 proteins ranging in size from 17 to 160 residues starting from known secondary structure. Predictions are obtained from global minimization of an empirical potential function followed by the application of a refined atomic overlap potential. The minimization strategy employed represents a variant of the Monte Carlo plus minimization scheme of Li and Scheraga applied to a reduced model of the protein chain. For all of the cases except beta-proteins larger than 75 residues, a native-like structure, usually 4-6 A root-mean-square deviation from the native, is located. For beta-proteins larger than 75 residues, the energy gap between native-like structures and the lowest energy structures produced in the simulation is large, so that low RMSD structures are not generated starting from an unfolded state. This is attributed to the lack of an explicit hydrogen bond term in the potential function, which we hypothesize is necessary to stabilize large assemblies of beta-strands.

Protein tertiary structure prediction using a branch and bound algorithm.

We report a new method for predicting protein tertiary structure from sequence and secondary structure information. The predictions result from global optimization of a potential energy function, including van der Waals, hydrophobic, and excluded volume terms. The optimization algorithm, which is based on the alphaBB method developed by Floudas and coworkers (Costas and Floudas, J Chem Phys 1994;100:1247-1261), uses a reduced model of the protein and is implemented in both distance and dihedral angle space, enabling a side-by-side comparison of methodologies. For a set of eight small proteins, representing the three basic types--all alpha, all beta, and mixed alpha/beta--the algorithm locates low-energy native-like structures (less than 6A root mean square deviation from the native coordinates) starting from an unfolded state. Serial and parallel implementations of this methodology are discussed.

A branch and bound algorithm for protein structure refinement from sparse NMR data sets.

We describe new methods for predicting protein tertiary structures to low resolution given the specification of secondary structure and a limited set of long-range NMR distance constraints. The NMR data sets are derived from a realistic protocol involving completely deuterated 15N and 13C-labeled samples. A global optimization method, based upon a modification of the alphaBB (branch and bound) algorithm of Floudas and co-workers, is employed to minimize an objective function combining the NMR distance restraints with a residue-based protein folding potential containing hydrophobicity, excluded volume, and van der Waals interactions. To assess the efficacy of the new methodology, results are compared with benchmark calculations performed via the X-PLOR program of Brünger and co-workers using standard distance geometry/molecular dynamics (DGMD) calculations. Seven mixed alpha/beta proteins are examined, up to a size of 183 residues, which our methods are able to treat with a relatively modest computational effort, considering the size of the conformational space. In all cases, our new approach provides substantial improvement in root-mean-square deviation from the native structure over the DGMD results; in many cases, the DGMD results are qualitatively in error, whereas the new method uniformly produces high quality low-resolution structures. The DGMD structures, for example, are systematically non-compact, which probably results from the lack of a hydrophobic term in the X-PLOR energy function. These results are highly encouraging as to the possibility of developing computational/NMR protocols for accelerating structure determination in larger proteins, where data sets are often underconstrained.