SCOP-Aided Fragment Assembly Protein Structure Prediction

Despite some limited success, computational biology has not been able to produce reliable results in the field of protein structure prediction. Although the fragment assembly approach has shown a lot of potential, it still requires substantial improvements. Not only are its predictions largely inaccurate whenever a protein exceeds 150 amino acids in length, but also, even for short targets, inconsistencies of the energy function associated with the enormous search space too often lead to the generation of erroneous conformations. Moreover, as it relies on the creation of a large number of decoys, it is highly computational expensive. Based on its secondary structure content, a protein can generally be classified into one of the standard structural classes, i.e. all-alpha, all-beta or alpha-beta. Since structural class prediction has reached a prominent accuracy, it is proposed to amend the standard pipeline of fragment-based methods by including some constraints on the template proteins from which fragments are extracted. Using Rosetta, a state-of-the-art fragment-based protein structure prediction package, the suggested customized method was evaluated on 67 former CASP targets ranging from 47 to 149 amino acids in length. Using SCOP-based structural class annotations, improvement of structure prediction performance is highly significant in terms of GDT (53 out of 67 targets show higher scores of 6.1% on average, p-value < 0.0005).