Re: [Psidev-ms-dev] possible charge states and etd

[Matthew C. <mat...@va...>]

Hi Chris,

I'm forwarding to psidev-ms since this is very relevant there.

In mzML, there is a special mechanism for storing a spectrum that is 
derived from other spectra: combined/merged spectra. I think with 
sufficient hackishness we could use the same mechanism to replace the 
original spectrum with multiple merged spectra that aren't really 
merged, but are just processed different ways and end up with different 
metadata (a hypothesized charge state instead of a possible charge 
state) and different binary data (different precursor peaks removed). 
The concept is reasonable, but the "merged=" id prefix should probably 
be augmented with "processed=" or something like that. The scanList 
would be the way that the processed version maps back to the original one.

-Matt


Chris Paulse wrote:
>
> For spectra where the precursor is either not confidently identified 
> as having a definitive charge state, or where two precursors with 
> different charge states co-fragment, it becomes a bit hard for me to 
> figure out what to do during the ms2 precursor removal step. There’s 
> an existing program called DTAGenerator that for cases where the 
> precursor charge is unknown, creates a separate spectrum record for 
> each assumed parent charge, and performs unfragmented precursor 
> removal for each hypothesis. This creates obvious problems for 
> downstream processing.
>
> On the other hand, it might not be the best idea to remove all charge 
> reduced precursors for all possible charge states in a single spectrum 
> before searching. The search task should have some way of generating 
> multiple hypotheses, and merging/filtering search results for the 
> given observation.
>
> I’m working with the Charge Prediction Machine program to identify 
> charge states for low resolution spectra – Zhi Sun wrote a perl script 
> that incorporates it into tpp. The idea would be to take the output of 
> this step (an mzXML file) and run msconvert on it. CPM does not 
> discriminate between indeterminate precursor charge and multi-charge 
> precursors. The author says that this is hard to do in practice 
> (probably because of lack of training data?). You might be aware that 
> Jimmy Eng has done some work to create an SVM equivalent of Charge 
> Prediction Machine, but this is still early in development.
>
> Do you see how this might impact the design of the precursor removal 
> spectrum filter?
>
> Thanks!
>
> Chris
>