Check non-flatness

[chmnk]

Some models predict non-flat sp3 rings as flat aromatic ones, as for some sets of atoms aromatic rings prevail in the training set. Would be nice to have a non-flatness check.

I'd be happy to commit this, but not sure what way would you prefer? Making a new file for non-flatness, making a new function, or just adding an option to choose between <= or >= operators to the check_flatness?

[maabuu]

Great idea. I welcome a PR on this. I would prefer the last option, that is adding an option to the flatness function to reuse it for non-flat systems as shown below. I am unsure about the threshold or the SMARTS here - maybe you have some example SMARTS or systems in mind?

Candidate non-flatness config for rings

  - name: "Ring non-flatness"
    function: "flatness"
    parameters:
      flat_systems: # list atoms which together should lie on plane as SMARTS matches
        nonaromatic_5_membered_rings_sp3: "[C,N&!a]1[C,N&!a][C,N&!a][C,N&!a][C,N&!a]1"
        nonaromatic_6_membered_rings_sp3: "[C,N&!a]1[C,N&!a][C,N&!a][C,N&!a][C,N&!a][C,N&!a]1"
      threshold_flatness: 0.25 # max distance in A to closest shared plane
      check_nonflat: True
    chosen_binary_test_output:
      - flatness_passes
    rename_outputs:
      num_systems_checked: number_aromatic_rings_checked
      num_systems_passed: number_aromatic_rings_pass
      max_distance: aromatic_ring_maximum_distance_from_plane
      flatness_passes: "Aromatic ring flatness"

Existing flatness config for rings

  - name: "Ring flatness"
    function: "flatness"
    parameters:
      flat_systems: # list atoms which together should lie on plane as SMARTS matches
        aromatic_5_membered_rings_sp2: "[ar5^2]1[ar5^2][ar5^2][ar5^2][ar5^2]1"
        aromatic_6_membered_rings_sp2: "[ar6^2]1[ar6^2][ar6^2][ar6^2][ar6^2][ar6^2]1"
      threshold_flatness: 0.25 # max distance in A to closest shared plane
    chosen_binary_test_output:
      - flatness_passes
    rename_outputs:
      num_systems_checked: number_aromatic_rings_checked
      num_systems_passed: number_aromatic_rings_pass
      max_distance: aromatic_ring_maximum_distance_from_plane
      flatness_passes: "Aromatic ring flatness"

[chmnk]

Thank you!

I am unsure about the threshold or the SMARTS here

the default >=0.1 threshold actually looks ok to me

I'd probably go for something like this, allowing for the nonflat systems with double bonds?

"nonaromatic_5_membered_rings": "[C,O,S,N&!a]~1[C,O,S,N&!a][C,O,S,N&!a][C,O,S,N&!a][C,O,S,N&!a]1",
"nonaromatic_6_membered_rings": "[C,O,S,N&!a]~1[C,O,S,N&!a][C,O,S,N&!a][C,O,S,N&!a][C,O,S,N&!a][C,O,S,N&!a]1",

I was also thinking about these:

"nonaromatic_6_membered_rings_db03": "[C,O,S,N&!a]~1[C,O,S,N&!a][C,O,S,N&!a]~[C,O,S,N&!a][C,O,S,N&!a][C,O,S,N&!a]1",
"nonaromatic_6_membered_rings_db02": "[C,O,S,N&!a]~1[C,O,S,N&!a][C,O,S,N&!a][C,O,S,N&!a]~[C,O,S,N&!a][C,O,S,N&!a]1",

but 'C1C=CN=CN1' turns out to be flat.