Lemke, O. and Keller, B.G. (2016) Densitybased cluster algorithms for the identification of core sets. Journal of Chemical Physics, 145 (164104).

PDF
4MB 
Official URL: http://dx.doi.org/10.1063/1.4965440
Abstract
The coreset approach is a discretization method for Markov state models of complex molecular dynamics. Coresets are disjoint metastable regions in the conformational space, which need to be known prior to the construction of the coreset model. We propose to use densitybased cluster algorithms to identify the cores. We compare three different densitybased cluster algorithms: the CNN, the DBSCAN and theJarvisPatrick algorithm. While the coreset models based on the CNN and DBSCAN clustering are wellconverged, constructing coremodels based on the JarvisPatrick clustering cannot be recommended. In a wellconverged coreset model, the number of core sets is up to an order of magnitude smaller than the number of states in a conventional Markov state model with comparable approximation error. Moreover, using the densitybased clustering one can extend the coreset method to systems which are not strongly metastable. This is important for the practical application of the coreset method because most biologically interesting systems are only marginally metastable. The key point is to perform a hierarchical densitybased clustering while monitoring the structure of metric matrix which appears in the coreset method. We test this approach on a moleculardynamics simulation of a highly exible 14residue peptide. The resulting coreset models have a high spatial resolution and can distinguish between conformationally similar yet chemically different structures, such as registershifted hairpin structures.
Item Type:  Article 

Subjects:  Mathematical and Computer Sciences > Mathematics > Applied Mathematics Biological Sciences > Biology 
ID Code:  1943 
Deposited By:  Ulrike Eickers 
Deposited On:  01 Sep 2016 15:33 
Last Modified:  03 Mar 2017 14:42 
Repository Staff Only: item control page