jmichel80
commented
4 years ago I haven't tested myself the effect of sanitizing the input. The partial ring match has proven useful for the Cresset implementation but could potentially be problematic in some instances.
We should expose these keywords but keep defaults to the current behaviour to facilitate testing as the best practices haven't been conclusively worked out.
It would be interesting to check if one protocol appears consistently better in terms of e.g. size of the MCSS.
Sent from my iPhone
On 1 Oct 2019, at 17:34, Lester Hedges notifications@github.com<mailto:notifications@github.com> wrote:
In figuring out the reason for some MCS mapping discrepancies between BioSimSpace and Cresset I have discovered that we do not sanitizehttp://www.rdkit.org/docs/RDKit_Book.html#molecular-sanitization molecules before performing the MCS match. This is because I followed the exact approach used by FESetup, where no sanitization is performed, e.g. the following which is taken from mutate/util.pyhttps://github.com/CCPBioSim/fesetup/blob/master/mutate/util.py:
mol1 = rdkit.Chem.MolFromMol2Block(mol2str_1, sanitize = False, removeHs = False)
The Cresset FEP pipeline uses a modified version of the open source LOMAPhttps://github.com/MobleyLab/Lomap to generate its MCS mapping. This too uses RDKit for the actual grunt work, however molecules are sanitized on load. Another difference is the use of completeRingsOnly=False when calling rdFMCS (we set to True).
Should we be sanitizing the molecules? Is there are reason why FESetup chose not to do this? (Perhaps it was unreliable at the time, or we want to apply our own chemical intuition to the generated MCS mapping.)
If we do want to sanitize, then it's a trivial change. However, I'll need to update all of the MCS tests since the mappings that are generated are slightly different.
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHubhttps://github.com/michellab/BioSimSpace/issues/124?email_source=notifications&email_token=ACZN3ZEG7G2WDJTMT25ZBLTQMN32ZA5CNFSM4I4L6O7KYY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4HO42YFA, or mute the threadhttps://github.com/notifications/unsubscribe-auth/ACZN3ZDIHG4A5LGZAQ5O5HLQMN32ZANCNFSM4I4L6O7A.
The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.
lohedges
In figuring out the reason for some MCS mapping discrepancies between BioSimSpace and Cresset I have discovered that we do not sanitize molecules before performing the MCS match. This is because I followed the exact approach used by FESetup, where no sanitization is performed, e.g. the following which is taken from mutate/util.py:
The Cresset FEP pipeline uses a modified version of the open source LOMAP to generate its MCS mapping. This too uses RDKit for the actual grunt work, however molecules are sanitized on load. Another difference is the use of
completeRingsOnly=Falsewhen callingrdFMCS(we set toTrue).Should we be sanitizing the molecules? Is there are reason why FESetup chose not to do this? (Perhaps it was unreliable at the time, or we want to apply our own chemical intuition to the generated MCS mapping.)
If we do want to sanitize, then it's a trivial change. However, I'll need to update all of the MCS tests since the mappings that are generated are slightly different.