Rank aggregation methods for the Web
Proceedings of the 10th international conference on World Wide Web
Proceedings of the thirty-ninth annual ACM symposium on Theory of computing
Aggregating inconsistent information: Ranking and clustering
Journal of the ACM (JACM)
Fixed-Parameter Algorithms for Kemeny Scores
AAIM '08 Proceedings of the 4th international conference on Algorithmic Aspects in Information and Management
The WEKA data mining software: an update
ACM SIGKDD Explorations Newsletter
Comparing Voronoi and Laguerre Tessellations in the Protein-Protein Docking Context
ISVD '09 Proceedings of the 2009 Sixth International Symposium on Voronoi Diagrams
Deterministic algorithms for rank aggregation and other ranking and clustering problems
WAOA'07 Proceedings of the 5th international conference on Approximation and online algorithms
Ultra-fast FFT protein docking on graphics processors
Bioinformatics
Computational Geometry: Theory and Applications
Hi-index | 0.00 |
Predicting the three-dimensional (3D) structures of macromolecular protein-protein complexes from the structures of individual partners (docking), is a major challenge for computational biology. Most docking algorithms use two largely independent stages. First, a fast sampling stage generates a large number (millions or even billions) of candidate conformations, then a scoring stage evaluates these conformations and extracts a small ensemble amongst which a good solution is assumed to exist. Several strategies have been proposed for this stage. However, correctly distinguishing and discarding false positives from the native biological interfaces remains a difficult task. Here, we introduce a new scoring algorithm based on learnt bootstrap aggregation ("bagging") models of protein shape complementarity. 3D Voronoi diagrams are used to describe and encode the surface shapes and physico-chemical properties of proteins. A bagging method based on Kendall-τ distances is then used to minimise the pairwise disagreements between the ranks of the elements obtained from several different bagging approaches. We apply this method to the protein docking problem using 51 protein complexes from the standard Protein Docking Benchmark. Overall, our approach improves in the ranks of near-native conformation and results in more biologically relevant predictions.