Improved estimation of structure predictor quality

We examined techniques to estimate the quality of a predicted protein structure based on prediction consensus. LGA is used to align the structure in question to the structures for the same protein predicted by different servers. We examine both static (e.g. averaging) and dynamic (e.g. support vector machine) methods for aggregating these distances on two datasets.We find that a constrained regression approach shows consistently good performance. Although it is not always the absolute best performing scheme, it is always performs on par with the best schemes across multiple datasets. The work presented here provides the basis for the construction of a regression model trained on data from existing structure prediction servers.The problem of predicting protein structure from amino acid sequence has yet to be fully solved, and experimentally determining protein structures requires extensive human input. Due to the relative ease of determining amino acid sequences, and the utility of structural information, the problem has attracted much attention. As the accuracy of protein structure prediction algorithms has greatly improved over the last ten years [1,2], the ability to precisely determine the quality of protein structure predictions has gained importance. In an attempt to motivate improvements in this area, the most recent session of the Critical Assessment of Structure Prediction (CASP) included a model quality assessment category [3]. CASP is A biennial competition http://predictioncenter.org webcite that has been organized to motivate advancements in the area of protein structure prediction. For this new category, given a putative structure, competitors were asked to submit a quality score between 0.0 and 1.0, or to assign an error estimate (in ？) to each residue of the structure. Twenty-eight groups submitted full structure quality estimates, and eight of those submitted per-residue error estimates.These eight groups assess structure quality using a variety of m