Methods of smoothing
This section will give you an insight into the method of paths smoothing that you should apply in order to smooth the inlet paths that you analyse. A trajectory of a solvent molecule inside a macromolecule’s core is usually very complicated and hard to follow. The molecule during passage through the protein core can visit different cavities and meanders of the protein interior. It can be simplified by smoothing in order to improve its visualisation and approximations of the central line of a particular tunnel used by the molecule. Later on, such approximation can be utilised for energy profile construction.
Current AQUA-DUCT version allows soft smoothing only, i.e. smoothing is used only for visualization purposes. During smoothing, the number of points (frames) in the smoothed path is the same as in a raw path. So the analysis of the smoothed path is done on the same number of points as in the raw path. However, for the purposes of visualization, some points in the path are removed. Only those points are removed which are aligned on the result smoothed path, so that no bias is introduced and removal of those points does not change the smoothed path’s appearance. The smoothing is also performed proportionally, i.e. in the end, the distance between individual points will more or less give information about the time it took the molecule to travel a given section. Studying smoothed paths can help to better understand solvent molecule’s routes during the simulation, their direction and sites where they are stuck for some time due to either attraction by protein’s amino acids or blocking by flexible parts of macromolecule’s interior.
Generally, there are four groups of smoothing methods applied in AQUA-DUCT:
- Window methods – window, dwin, awin – methods based on producing mean or median values in frames (windows)
along the path.
- Maximal step method – mss – method which creates an average of coordinates in a given range of path steps.
- Combined methods – window_mss, dwin_mss, awin_mss – a combination of window- and mss-based methods. First,
mss smoothing is performed and then window smoothing.
- Savitzky-Golay filter – savgol – a distinct method based on a process known as convolution, which fits a polynomial
to subsets of points along the path.