The authors describe an interactive system for understanding the structure of tunnels (a potential path for a ligand) through a protein, as computed from molecular dynamics simulations. Mapping the complex 4D shape of the tunnel, and the dozens of amino acids involved, to a visually accessible 2D summary of plots and ranked amino acids, involved numerous design decisions that can be appreciated from an algebraic standpoint.
Good visualization design requires thoughtful decisions about what to show, and how to show it. The visual-data correspondence principle suggests that the possible changes to the data (the alpha) most relevant to a given investigation should map to visual changes (the omega) that are the most legible. For understanding how a ligand may access part of a protein via a tunnel, the width of the tunnel is the most important quantity, so the authors map this to position along a common scale, the highest-ranked visual affordance according to Cleveland and McGill (1984). The upper pane of the interface shows plots of tunnel width as a function of depth, one plot per simulation time step.
All bottlenecks (minima of width), and the stability of the bottlenecks, become immediately visible, because other confounding degrees of freedom in 3D shape, orientation, and location, have been projected out. The change in the tunnel shape before and after a single amino acid substitution (alpha), is thereby made clearly visible by a large change in the plots (omega), shown in the change between Figures 11 and 12. The simultaneous overlay of plots for all different time-steps does create a confuser (a problem for the unambiguity principle), but this can be removed by a univarate colormap of time (Fig. 1).
Investigators know which amino acids along the path are especially important to target thanks to the second lower pane of the interface, displaying where (along the tunnel length) individual amino acids are closest to the tunnel centerline, as ranked by biochemically relevant variables, such as hydrophobicity (repelled from water). Changes in amino acids that cause a different in these variables will result in a visibly different ranking, displayed in a linear vertical ordering, in keeping with visual-data correspondence. Hydrophobicity is visualized with a double-ended colormap of opponent colors, ranging from hydrophilic (magenta) to hydrophobic (purple); diverging colormaps for signed quantities are another example of visual-data correspondence.