Towards a better measure of protein 3D model quality

Advances in protein three-dimensional structure prediction depend strongly on the ability to measure the quality of a protein model. One of the best single measures of the model quality is the area under the graph of the so-called "GDT function" that assigns to each distance cutoff theta the percentage of residues in the model structure that can be superimposed at distance <=theta from the corresponding residues in the experimental structure. In lieu of a method capable of computing this measure, researchers often resort to its approximation GDT_TS. In this short paper we present the first theoretical framework for the development of an algorithm capable of computing near-optimal estimates of the area under the GDT curve of a protein model in polynomial time.