10 likes | 134 Views
Protein Flexibility Modeling Using Kernel Methods X ue-wen Chen and Jeremy Chen Bioinformatics and Computational Life Science Laboratory, ITTC Department of Computer Science, The University of Kansas, Lawrence, KS 66045 E‑mail: xwchen@ku.edu. 3. Relevant Work Collective mode based
E N D
Protein Flexibility Modeling Using Kernel MethodsXue-wen Chen and Jeremy ChenBioinformatics and Computational Life Science Laboratory, ITTC Department of Computer Science, The University of Kansas, Lawrence, KS 66045 E‑mail: xwchen@ku.edu • 3. Relevant Work • Collective mode based • decompose the protein flexibility into collective modes • Singular value decomposition based • analyze MD trajectories of atomic fluctuations • Principal component analysis based • reduce the high dimensionality • All the methods are linear in nature; cannot characterize the nonlinearity embedded in protein flexibility • Need nonlinear modeling for protein flexibility – kernel principal component analysis based modeling 2. Example: Drug Discovery Molecular Docking – Finding candidate drug binds to the active site of protein which disrupts the function of protein. One problem -- some proteins undergo conformation changes during binding (induce fit) Current docking methods ignore this fact due to computational complexity. Only can model rigid protein structures, due to the high degrees of freedom • 1. Introduction • Proteins play an essential role in nearly all cell functions. • 3D structures have the most direct relationship with protein functions • Protein flexibility: Protein dynamic changes in conformation; an important link between protein structure and function • Analysis of a single protein conformation provides certain information about protein operations • Understanding the conformational changes is essential in discovering many biological process Library Molecular Docking Target Protein Courtesy: Amarda, Shehu 6. Conclusions Conclusion: with KPCA, few principal components are needed to characterize the conformation changes. Active Site: Entire System: Acknowledgement: We thank NIH Grant P20 RR17708 from the Institutional Development Award (IDeA) Program of the National Center for Research Resources. 5. Results Protein data: HIV-1 Protease 4. Kernel Principal Component Analysis RASMOL generated image for HIV Protease (4HVP) at different angles with ligand binded (courtesy: Wong Benton) Generating conformational data using NAMD 4HVP solved in water box in different angles