A population-based conformational optimal algorithm using replica-exchange in ab-initio protein structure prediction

Referring to ab-initio protein structure prediction, a replica exchange based differential evolution searching method (REDE) was proposed. Firstly, employing fragment assembly technique to reduce the dimension of protein conformational space effectively, thus avoiding the entropy effect in searching. Additionally, every replica layer equipped with a conformation population, updating the population by using differential evolution algorithm, to ensure diversity of population in replicas. Moreover, it can enhance the capacity of algorithm to jump out of local minimum. The result of 10 test proteins show that the mentioned algorithm can rapidly and effectively sample from the conformational space and generated high-resolution near-native protein conformations.