ICESS'04 Proceedings of the First international conference on Embedded Software and Systems
High performance 3D convolution for protein docking on IBM blue gene
ISPA'07 Proceedings of the 5th international conference on Parallel and Distributed Processing and Applications
| Hi-index | 0.00 |
We have developed a system, which we call affinity evaluation and prediction (AEP) system, to evaluate and predict the partners in protein-protein interaction (PPI) by using a statistical method for calculated docking scores. The complex-protein structures obtained in shape complementary evaluation are selected by a newly developed clustering method called grouping. Our previous experiments showed that the AEP system has a tendency to give different accuracy depending on biologically significant protein dataset and data scale (20×20= 400 protein pairs). In this study, we set a data scale (54×54=2916 protein pairs) including 54 biologically significant complexes. As a result of receiver operating characteristics (ROC) analysis, the AEP system obtained 55.6% sensitivity (=recall), 70.6% specificity, 3.44% precision, 70.3% accuracy, 6.48% F-measure and an area under the curve (AUC) of 0.655. The prediction accuracy of F-measure was about 1.82 times higher than that of a random sampling (F-measurerandom=3.57%). By optimizing the grouping procedure, the AEP system successfully predicted 30 protein pairs (among 54 pairs) that were biologically significant.