Finding Semirigid Domains in Biomolecules by Clustering Pair-Distance Variations
Figure 10
Time dependence of target function, evaluated for 50 subset-trajectories and different number of clusters. The total trajectory of 250 ns was split into 50 consecutive subset-trajectories and clustering performed for each of them, prescribing the number of clusters between 2 and 6 (see coloured legend). For each clustering, the resulting averaged STDDV is plotted against the vertical axis. Averaged STDDV have been multiplied by (as in Figure 7) to make them comparable between different numbers of clusters. Generally, more clusters entail smaller cluster size and thus reduce total motility captured in clusters. Variability of STDDV between subtrajectories illustrates the dynamical character of clustering and its dependence on the phase space sampling stage, however, with a pronounced convergence tendency of local values.
