A Population-Based Approach for Diversified Protein Loop Structure Sampling

Protein loop structure modeling is regarded as a mini protein folding problem with significant scientific importance. Efficiently sampling the loop conformation space is a key step to computationally obtain accurate loop structure models. Due to the large size of the conformation space and the complication of the scoring functions describing protein energy, it is difficult to obtain broad, diverse coverage of the loop conformations with low energy (score). In this article, we present a new population-based approach to sample the backbone conformations of protein loops. The main advantage of the population-based approaches is that various selection schemes can be applied to enforce the conformations in a population to satisfy certain constraints. In our sampling approach, conformations are generated in the dihedral angles (φ,ψ)space and the Differential Evolution (DE) method is employed to implement dihedral angle crossover for generating new conformations. A diversity selection scheme is applied to achieve diversified sampling. Using a narrowing gap selection scheme, decoys satisfying loop closure condition are obtained by gradually eliminating conformations with large terminal gaps in a population. Our computational results on modeling long loop targets have shown diverse and broad coverage of the loop conformation space, which leads to consistently reaching the native-like decoys in the sampling process.