| Popis: |
This chapter reviews automated histogram filtering (AHF), a recently developed approach to cluster analysis, and describes two applications of this methodology, illustrating its power as a tool for analysis of biomolecular simulation data. The first is to cluster low-energy conformations of a model protein, generated by Monte Carlo-simulated annealing. From ensuing cluster memberships, we arrive at distance constraints to serve as input to distance-geometry software. Additional low-energy structures are subsequently generated by the latter software with remarkable efficiency. The second application is to cluster molecular dynamics snapshots of an all-atom model of the response regulator protein Spo0F of Bacillus subtilis . Two clusters are uncovered, reflecting our having sampled two distinct basins on the potential energy surface. We subject snapshots for each cluster to AHF cluster analysis, and perform an orthogonal rotation of the dominant so-called ‘principal components’. We monitor the loadings of each interatomic distance variable along the rotated principal components. To visualize our results, we construct squared-loadings plots, revealing secondary structural elements that undergo concerted motions along so-called ‘collective coordinates’. Dominant collective coordinates are found to be associated with contiguous stretches of amino acid residues implicated in the biological functioning of Spo0F. |
