Mismatch between adduct and ionmode after filtering pipeline

[niekdejonge]

@Philipbear reached out to show some mistakes in the cleaned pipeline.

Some adduct forms are not consistent with ion mode & charge. E.g., CCMSLIB00000001723 (below). In this case, the adduct looks like to be [M]+. I guess this is raised due to the tiny mass of electron. It would be good if adduct, ion mode and charge are consistent among each other.

One more example is CCMSLIB00000007028. Adduct is [M+K-2H]-, yet it is in positive ion mode. I think in case it would be [M+K-H2]+.

To solve this I propose the following two solutions:

• Adding a filter that requires charge, ionmode and adduct charge to match. We already have many repair filters that derive the charge or ionmode from adducts when this field is missing, but we should also remove spectra that have a mismatch between them.
• We currently allow the repair_adduct_based_on_smiles to repair adducts even if this does not match the Ionmode for this spectrum. This should not be the case, since there is an increased risk of repairing adducts in the wrong way. So we should only searched in adducts that match the ionmode given for the spectrum.

[hechth]

How do you repair adduct based on SMILES? Do you expect the SMILES to include the adduct? I thought the SMILES should normally represent the whole molecule?

[niekdejonge]

We compare the smiles to the precursor mz and if the mass difference matches a common adduct the adduct and parent mass are updated accordingly. The smiles are indeed expected to represent the whole molecule.