Prediction of protein structures using a map-reduce Hadoop framework based simulated annealing algorithm

The accomplishment of molecular functions depends on protein tertiary structures. The development of protein structure prediction algorithms and tools is essential for proteomics study. Among the existing developed prediction algorithms, simulated annealing (SA) is extensively used to predict protein structures. However, SA has the incent disadvantages of computing time consuming and local minimum convergence problem. With the application of the cloud computing technique such as Apache Hadoop in bioinformatics research area, we combined SA algorithms to predict protein structures onto the Hadoop parallel computing platform. We applied this platform to predict the protein structures for a public protein dataset. The experimental results show that our platform provides a better and feasible solution for the protein structure prediction compared with an individual computation node.