Feat/update vdw radii #745
Conversation
CodSpeed Performance ReportMerging #745 will not alter performanceComparing Summary
|
|
Not sure what's happening on the CI side here as I only changed a dictionary -- maybe a glitch? |
There was a problem hiding this comment.
Thanks for the update. I only left two minor comments, otherwise this looks good!
I will also add the new reference (Mantina2009) to the Biotite Zotero library.
Regarding the failing test: It fails, because the reference SASA values were computed with mdtraj using the old radii. To update the reference values to the new radii, just rerun sasa.py (requires mdtraj). I can support you here, if you need some help.
| @@ -83,7 +83,8 @@ def sasa(array, float probe_radius=1.4, np.ndarray atom_filter=None, | |||
| - **Single** - A set, which uses a defined VdW radius for | |||
| every single atom, therefore hydrogen atoms are required | |||
| in the model (e.g. NMR elucidated structures). | |||
| :footcite:`Bondi1964` | |||
| Values for main group elements are taken from :footcite:`Mantina2009`, | |||
| and for relevant transition metals from the RDKit. | |||
There was a problem hiding this comment.
Could you also add a citation to the relevant RDKit paper here?
| and for relevant transition metals from the RDKit. | |
| and for relevant transition metals from *RDKit*. |
| ("I", 1, 0) : 1.98, # Taken from _SINGLE_RADII | ||
| } | ||
|
|
||
| _SINGLE_RADII = { | ||
| "H": 1.20, | ||
| _SINGLE_ELEMENT_VDW_RADII = { |
There was a problem hiding this comment.
Just an optional nit: I think single atom is more appropriate than single element here.
| _SINGLE_ELEMENT_VDW_RADII = { | |
| _SINGLE_ATOM_VDW_RADII = { |
I had a detailed look into the latest open access vdW radii, especially for elements that were not previously represented in biotite. I updated the values with recent authoritative sources, which should now allow SASA computations for almost all structures. I also did a comparison against non-open access reference values which we cannot use here, but made sure that open access values we use are close to (usually <0.05A) those reference values.