Dihedral Clustering of MD simulations
For structure prediction bioactive poeptides, we invented an analysis and mapping algorithm for multi-dimensional dihedral angle time-series data from molecular dynamics trajectories. It is based on a procedure for clustering followed by a Sammon-mapping of the cluster centers. The procedure proves to be superior to Cartesian clustering, as it is capable to determine small scale local conformational changes that are often not distinguisheable in Cartesian clusterings. Furthermore, our cluster centers represent phyically meaningful conformations, enabling for a better analysis of essential molecular motions, e.g., by principal component analysis. The method will be useful for the analysis of folding properties and thus for the predction of peptide structures of natural substances.
The picture on the right shows density plots as contours of dihedral angles found in a 10ns MD trajectory of Met-Enkephalin. The superimposed elipses are from the newly derived clustering strategy.
published in Journal of Computational Chemistry, (2009)
Co-operation:
Martin Zacharias, TU-Munich
Molecular Dynamics Simulation
A 4 nanosecond molecular dynamics simulation of a hexa peptide in a 40A waterbox at 310 K is shown here. The MD-program used is NAMD, preparations were done with VMD. The simulation was computed using 2 CPUs of an Opteron bioinformatics cluster server. It took 2 days to compute.
A cluster analysis into 13 main conformations within a 4x4 Self Organizing Map is performed, resulting in two main conformers, the C-wiggle and the S-wobble . The original trajectory was split into frames that best fitted one of the SOM-units. The SOM-input data consisted of the 15 most relevant dihedral angles of the hexa-peptide trajectory. For this analysis, VMD-scripts trajsort/mkpic/overl were written in tcl. For the SOM, som_pak-3.1 was modified for periodicity of the dihedral agles.
The pictures on the left show the main conformers as superimposed from a selected sub-set of SOM-clusters.
Co-operations:
M. Köck, Sv. Meyer and G. Schmidt (AWI)
Bundles from dihedral SOM-clustering
multi44
All SOM units rotating in a movie