A physics-based scoring function for protein structural decoys: Dynamic testing on targets of CASP-ROLL

Most successful structure prediction strategies use knowledge-based functions for global optimization, in spite of their intrinsic limited potential to create new folds, while physics-based approaches are often employed only during structure refinement steps. We here propose a physics-based scoring potential intended to perform global searches of the conformational space. We introduce a dynamic test to evaluate the discrimination power of our function, and compare it with predictions of targets from the CASP-ROLL competition. Results demonstrate that this dynamic test is able to generate 3D models which outrank 59% (according GDT-TS score) of models generated with ab initio structure prediction servers.Yasser B. Ruiz-Blanco, Yamila García and Clivia M. Sotomayor-Torres acknowledge the partial support of the Spanish MINECO grant MAT-2012-31392 (TAPHOR). Yasser B. Ruiz-Blanco acknowledges the ELAP scholarship to conduct research at Carleton University. Yovani Marrero-Ponce thanks the ‘International Professor’ program for a fellowship to work at Cartagena University in 2013–2014.Peer Reviewe