wrong formula in fragment spectrum and rare NaNs about sirius HOT 3 CLOSED

1. What is the precursorFormula column supposed to be in the formula_candidates.tsv file? From the past I know that this column is always equal to molecularFormula when having [M+H]+ adducts. Is it simply always the deprotonated form?

The precursor is the adduct that is selected for MS/MS. Hence, molecularFormula is the neutral molecular formula of the compound and percursorFormula the neutral formula of the measured adduct (in your case the [M+NH4]+).

2. Bug: the medianMassError from formula_candidates.tsv gives a NaN here :(

The medianMassError is a measure of the fragments in MS/MS. In your minimal example the MS/MS only contains one peak with m/z greater than the precursor mass. Therefore, the calculated fragmentation tree contains no fragments that can be used to calculate a median mass error. We might adjust the naming of this column to make this more intuitive.

3. The file name of the spectrum file seems to match that of precursorFormula from the candidates file, is that always the case? Again, when using [M+H]+ this equals that of the molecularFormula.

Yes, we name the spectrum and fragmentation tree files based on the precursorFormula and adduct to indicate the adduct that was measured by MS/MS.

4. The ionization column in the spectrum file suddenly says [M + H]+.

We differentiate "ionization" and "adduct". Ionizations are simple ion modes such as [M+H]+, [M+K]+ or [M+Na]+. And adduct might include another adduct fragment or in-source fragment ([M+NH3+H]+ or [M-H2O+H]+). Simply looking at the MS/MS, it is not obvious to differentiate if an MS/MS spectrum was produced by [H5NO5+H]+ or [H2O5+NH3+H]+. Because of that, we only considered the simple ion modes in the past. And currently, we still annotate the peaks in the spectrum file in this way. But since we now include feature detection in our workflow, we can estimate adducts of the precursor based on the MS1 more frequently. And based on the precursor adduct we can also make suggestions of the fragment adducts.
We had a discussion on that yesterday and decided that we will also annotate peaks in the spectrum file with a possible adduct type in cases where we estimate one based on the fragmentation trees. However, these adducts will always be suggestions since e.g. fragments [C6H15NO6+H]+ and [C6H12O6+NH3+H]+ produce the exact same m/z.

I hope this clarification helps.

Best,
Marcus

from sirius.

I hope the answers helped to resolve your questions. I will close this issue for now.

from sirius.

Hi Marcus,

Many thanks for all the explanations, that clarifies it well :-)

from sirius.