Background: Many functionally important proteins in a cell form complexes with multiple chains. Therefore,\r\ncomputational prediction of multiple protein complexes is an important task in bioinformatics. In the development\r\nof multiple protein docking methods, it is important to establish a metric for evaluating prediction results in a\r\nreasonable and practical fashion. However, since there are only few works done in developing methods for\r\nmultiple protein docking, there is no study that investigates how accurate structural models of multiple protein\r\ncomplexes should be to allow scientists to gain biological insights.\r\nMethods: We generated a series of predicted models (decoys) of various accuracies by our multiple protein\r\ndocking pipeline, Multi-LZerD, for three multi-chain complexes with 3, 4, and 6 chains. We analyzed the decoys in\r\nterms of the number of correctly predicted pair conformations in the decoys.\r\nResults and conclusion: We found that pairs of chains with the correct mutual orientation exist even in the\r\ndecoys with a large overall root mean square deviation (RMSD) to the native. Therefore, in addition to a global\r\nstructure similarity measure, such as the global RMSD, the quality of models for multiple chain complexes can be\r\nbetter evaluated by using the local measurement, the number of chain pairs with correct mutual orientation. We\r\ntermed the fraction of correctly predicted pairs (RMSD at the interface of less than 4.0Ã?â?¦) as fpair and propose to\r\nuse it for evaluation of the accuracy of multiple protein docking.
Loading....