Leaflet sorter speed up and shell selection bug fix

[JesseSandberg]

Description

When running polarDensityBin the leaflet sorter would run on all lipids in the system, even those outside of the zone of analysis. After this PR, the user can select whether leaflet sorting will only occur for lipids within Rmax or for all lipids.

In the process, we discovered a bug (issue #104). If the user specified an Rmax that was evenly divisible by dr, then polarDensityBin would analyze past Rmax by one shell. If the user specified an Rmax that was not evenly divisible by dr, then polarDensityBin would happily accept that choice without balking (but density enrichment analysis would be affected in the outermost shell downstream).

We also found that expected density was being measured on start_frame rather than end_frame (issue #107).

We also found a minor error when no lipids are detected within the shell (issue #110).

After this PR, shells will no longer exceed Rmax when Rmax is evenly divisible by dr. Additionally, polarDensityBin will give an error message and quit if Rmax is not evenly divisible by dr. This fix is agnostic to whether the user uses ints or floats in their config file. Expected density will be measured on end_frame.

Implementation of auto-rescaling dr (issue #105) will occur in a future PR

Usage Changes

• User has access to restrict_leaflet_sorter_to_Rmax parameter (default is 0).
• User will receive an error message if they specify Rmax not evenly divisible by dr.

Todos

Notable points that this PR has either accomplished or will accomplish.

• Leaflet sorting can either sort all lipids or only sort lipids within Rmax
• If Rmax is evenly divisible by dr, shells will no longer exceed Rmax
• Expected density measured on end_frame
• User receives error message f Rmax is not evenly divisible by dr.
• User receives error message if restrict_leaflet_sorter_to_Rmax is not 1 or 0 (default is 0)
• Bug fix: empty shell no longer causes error.

Pre-Review checklist (PR maker)

• Ensure no run-time errors
• Test with variety of Rmax and dr options
• Test with variety of restrict_leaflet_sorter_to_Rmax options
• Run same analysis twice - once prior to change and once after change - and then vimdiff the outputs to ensure they are identical
• Run same analysis twice - once prior to change and once after change - and ensure that resulting delta Gsite is the same at the very end (perform end-to-end testing)
• Include test in the DTA_Testing repo (here)

Review checklist (Reviewer)

• No missed "low-hanging fruit" that would substantially aid readability.
• Any "high-hanging" or "rotten" fruit is added to the issues list.
• I understand what the changes are doing and how
• I understand the motivation for this PR (the PR itself is appropriately documented)

Status

• Ready for review

[JesseSandberg]

closed because I found aberrant behavior with lipids that were only partially enclosed in the shell created by Rmax

[JesseSandberg]

Note from convo today: This PR will be reviewed when I have completely replicated analysis on the nachr project from end to end with this branch and gotten the same results out as previously.

[JesseSandberg]

As discussed in nougat meeting, I have re-run a DTA calculation from a separate project and confirmed that the site delta G does not change when these changes are made.

[JesseSandberg]

As discussed in the lab, I have created a separate repo - DTA_Testing - that now houses any testing notebooks and materials created.

[gbrannigan]

@JesseSandberg

This code is more complicated than it needs to be. Do a simple (i.e. double($Rmax) % $dr) modulo to get the remainder, and then compare the remainder against a small tolerance (not 0!).

Do not make the tolerance a magic number.

[JesseSandberg]

Code has been modified per discussion on slack. Thank you!

[gbrannigan]

I have left two comments, please address.