一种基于关联图的蛋白质结构预测改进算法

采用一种改进的COMAR （Contact Map Reconstruction）算法求解基于关联图的蛋白质结构预测问题.根据蛋白质关联图和先验知识，并以半随机的方式生成距离信息，根据距离信息得到蛋白质的坐标，并通过坐标修正和摄动，使得重构结构的关联图与给定的关联图相一致.结果表明，阈值较大的关联图所重建的结构较好，与原COMAR算法相比，在相同的迭代次数下，改进的COMAR算法的精度较高.

An Improved Algorithm for Protein Structure Prediction Based on Contact Map

An improved contact map reconstruction(COMAR) algorithm was introduced to reconstruct 3D structures from protein contact maps. After guessing a possible set of distances consistant with contact map, the proposed algorithm generates a random initial set of 3D coordinates, and then iteratively applies two local correction and perturbation techniques to the current set of coordinates. The results show that when the threshold of the contact map computation is increased, the reconstructed structure is more similar to the native structure. In addition, the RMSD computed by improved COMAR is lower than COMAR in the same case.