EVA: Large-scale analysis of secondary structure prediction

Abstract

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/.

Murine apolipoprotein serum amyloid A in solution forms a hexamer containing a central channel

“…The secondary (2°) structure of SAA2.2 was predicted by using several algorithms from the ProteinPredict web server (http:͞͞cubic.bioc.columbia.edu͞predictprotein). These methods are Ϸ76% accurate (25), and all gave similar results. In Fig.…”

Protein Structure Prediction using String Kernels

“…The neighbor-based approaches [53,14,23] predict the secondary structure by identifying a set of similar sequence-fragments with known secondary structure; the model-based approaches [49,22,44,42], employ sophisticated machine learning techniques to learn a predictive model trained on sequences of known structure; whereas the meta-predictor-based approaches [12,41] predict based on a combination of the results of various different neighbor and/or model-based techniques. The near real-time evaluation of many of these methods performed by the EVA server [48] shows that the model-based approaches tend to produce statistically better results than the neighbor-based schemes, which are further improved by some of the more recently developed meta-predictor-based approaches [41].…”

Bayesian network multi-classifiers for protein secondary structure prediction

“…• EVA1, ..., EVA6: 6 novel test sets are provided by the datasets available from the real-time evaluation experiment called EVA [10], which compares a number of prediction servers on a regular basis using the sequences deposited in the PDB every week. In particular we have used all the datasets labelled "common1" to "common6" published on 19/10/2002.…”

“…The most representative dataset designed by different groups are HS1771, CB513, RS126 as well as the six data sets gathered by EVA project [10]. From all of them only HS1771 has been used for the training phase of a multi-classifier.…”

“…We have used nine main datasets to train and test our approach: a training set, the PDB SELECT list [7] of March 2002 (HS1771), and eight test sets (RS126 [8], CB513 [9] and 6 different datasets from the EVA project [10]). …”