Skip to main content
Topic: annotation of doubly charged metabolites (Read 14084 times) previous topic - next topic

annotation of doubly charged metabolites

Hi,

on which basis does the CAMERA package determine the charge of a metabolite? I have got the impression that it does not always correctly determine the charge of metabolites. Does this heavily rely on matching of the EICs of isotopes? I have attached an excel sheet of an annotated diffreport of a metabolite group of which I think the annotation of the charge was incorrect. The data has been generated using XCMS online.

The annotation settings where:
annotation settings: isotopes
m/z absolute error: 0.015
ppm: 5


Koen

Re: annotation of doubly charged metabolites

Reply #1
Hi Koen,

the determination of the isotope charge is a separated step in the annotation pipeline.
Therefore, it relies on the mass differences within a already grouped pseudospectra (so grouping retention time or EICs).
For an example with three peaks: 270.0, 270.5, 271.0 m/z

Here 270.0 would get a [M]2+ annotation, 270.5 a [M+1]2+ and 271.0 a [M+2]2+.
Another possible annotation is 270 as [M]+ and 271 as [M+1]+, but since the first solution explains more peaks we decide for the majority.

Additionally we have some constrains to filter false positive. For example we estimate the minimum and maximum number of carbon atoms from the monoisotopic peak
and calculate a intensity range for the first isotopic peak.
Hope this helps a bit.

I can take a look into your example, but it seems the attachment is missing.

Carsten

Re: annotation of doubly charged metabolites

Reply #2
Hi Carsten,

Apparently it is not possible to upload xls or csv files. so I have copied the data underneath.

   name    fold                   pvalue   qvalue                   fscore   mzmed                   rtmed                   isotopes             pcgroup
1   M542T30    12.24798672   1.40E-05   0.00771968   210.5   542.2727759   29.93576667   [515][M]+             166
2   M543T30_2 12.27611853   1.81E-05   0.00771968   452   543.2755869   29.93576667   [515][M+1]+        166
3   M543T30_1 13.97008932   2.08E-05   0.00771968   38   542.7741446   29.97831917   [395][M]+             84
9   M544T30    16.55803621   5.77E-05   0.00771968   159   543.7769112   30.02087167   [395][M+1]+        84
15   M1086T30    16.08646843   9.54E-05   0.00950582   402.5   1085.553926   29.93576667                166



thanks a lot for your help!
Koen

Re: annotation of doubly charged metabolites

Reply #3
The last column pcgroup describes the compound or pseudospectrum index.
Means here that peak 1, 2 and 15 belongs after CAMERAs grouping to compound 166 and peak 3,9 to 84 whatsoever both compound are.
The isotope detection searches in both groups separately, which leads to the annotation of two single charged isotope clusters.
So if you are sure about the double charged metabolite, then the grouping parameter was to strict. Decrease the EIC correlation threshold to get all peaks in one group.

If not and without the intensity values it's hard to say and perhaps unlikely, but another possibility could be that we have here two co-eluting compounds with a mass difference of 0.5 m/z.
I'm only asking because from the very early retention time it seems like the injection peak, where you normally have a bunch of co-eluting compounds.

Carsten

Re: annotation of doubly charged metabolites

Reply #4
Hi Carsten,

thanks for your help. These compounds do not elute early, the retention times given are in minutes. So I will try to tweak the EIC correlation settings, but I am not sure whether this can be done within XCMS online.

Koen

Re: annotation of doubly charged metabolites

Reply #5
Hi,

I have a question along the same lines.  Does CAMERA use the intensity of the masses when assigning an isotope group, or just the the masses/retention time?  I have encountered a situation where the M+ is assigned to a mass that is 10000x less intense than the real M+1 of a known compound.  It may be an elimination product of the parent M+H that co-elutes or an artifact.  A picture:



Is there any way to prevent wrongly assigning isotopes by considering peak intensity? I'm not sure if this is in the scope of what CAMERA does.

Thanks, I hope this is clear as mud!

Kraig Worrall

Re: annotation of doubly charged metabolites

Reply #6
Hi,

so far CAMERA uses an intensity check only for the M+1 peak by estimating the number of C atoms in a molecule (min, max range).
The M+2 peak is more challenging without knowing the molecular formula, because more elements (e.g. O, S) contribute.
If your sample contains halogen atoms it's way more difficult.

But I see your point to get a maximum intensity / ratio estimation to resolve wrong annotation like in your example.
We will discuss this and are working on a solution, without expanding the parameter list too much.

Carsten

Re: annotation of doubly charged metabolites

Reply #7
Thanks Carsten!

I appreciate your support

Re: annotation of doubly charged metabolites

Reply #8
I have a similar question regarding annotation of doubly charged metabolites. Below I have an example of two pcgroup's where camera has identified them as [M]2- due to the fact that there is a 0.5 da mass difference between features.

[attachment=2:2tij0bq6]peaktable.PNG[/attachment:2tij0bq6]

In both cases I believe this is incorrect because when I look at the actual raw data each peak is actually 1.0 da apart
[attachment=1:2tij0bq6]806_isotope.PNG[/attachment:2tij0bq6]

Based on some of the comments above I'm guessing, this is likely due to a grouping and over splitting features based on grouping parameters  that are too strict? We used ( xset3 <-group(xset2, method="density",bw=5, mzwid=0.05, minfrac=0.1))
which parameter is causing this over splitting? bw of 5 or mzwid=0.05?

Similarly,  in the above peak table pcgroup 4 also has a similar issue where features are separated by 0.5 da intervals, however, the raw data does not show this.

[attachment=0:2tij0bq6]1384_isotope.PNG[/attachment:2tij0bq6]
 
This case seems a bit different becuase the mass differences aren't exactly 1.0 da (see below). Why would these masses not be exactly 1.0? My initial thought is that it is due to the fact that these features are from a cluster [3M-2H]2- but I'm not exactly sure how that would change the mzdif. can someone explain this?

mass diff of [mz]2- to [mz+1]2- = 0.7318
mass diff of [mz+1]2- to [mz+2]2- = 0.9445
mass diff of [mz+2]2- to [mz+3]2- = 0.9631


Thanks in advance any guidance or advice!