A Multi-Template Multi-Model Combination Approach to Template-Based Modeling Jianlin Cheng Computer Science Department & Informatics Institute University of Missouri, Columbia, MO, USA
5. Combination & Refinement (2-3%) 1. Template Ranking 2. Multiple-Template Combination Alignments Combination Query-Template 1 MAR-TCRK-EGAP-WY… Y-R-MH-R-DGM-MWT… TAKMTHK-DEGFG-YW… MARTCRKEGAP-WY… Y-RMH-RDGM-MWT… Input Query . MARTCRKE… Query-Template 2 MAR-TCRK-EGAPWY… TAKMTHK-DEGFGYW… . . 4. Evaluation 5. Combination & Refinement (2-3%) 3. Model Generation Models Generator Output CASP8 Server Models
Traditional Model Selection Single-Model Evaluation Clustering / Consensus Approach
Global-Local Model Combination CASP8 Models Rank models by GDT-TS scores predicted by ModelEvaluator …… . Put relatively good, but not the best models at the top
Global-Local Model Combination Structure comparison by TM-Score . . Select top 5 models as seed models Identify similar models or fragments Retain top 50% models
Global-Local Model Combination Globally similar models Locally similar model fragments Combination and iterative modeling by Modeller Side chain rebuilt by SCWRL.
Some High-Quality Predictions GDT=0.90 T0426 GDT=0.97 T0432 GDT=0.92 T0458 GDT=0.97 Orange: structure; Green: model H-Bonds are well predicted.
Conclusions Iterative modeling and averaging improve side-chain placement, geometry, and H-Bonds Combining multiple good similar models can produce a model better than the top ranked model Combined models are at least as good as centroids and have no steric clashes
Acknowledgements CASP8 organizers and assessors CASP8 participants MU colleagues: Dong Xu, Toni Kazic My group: Zheng Wang Allison Tegge Xin Deng