problem with mzClust

Topic: problem with mzClust (Read 4949 times) previous topic - next topic

problem with mzClust

September 19, 2012, 05:59:08 AM

Hi to all,

I have some problems comparing group.density and group.mzClust methods or maybe I didn't well understand th way group.mzClust works.
Indeed, when I group according to density method the diffreport is well composed of mz and rt data while with mzClust rtmed, rtmin and rtmax have only -1 value for all extracted ions.
is it normal ?

I point out that I work on HPLC-QTOF data.

here below you can find the scripts I used:

group.density

Code: [Select]

xset <- xcmsSet(method="matchedFilter", fwhm=10, snthresh=3, nSlaves=4)
xsetb <- group(xset, minfrac = 0.75, bw = 10)
xset2 <- retcor(xsetb, family="s", plottype="m")
xset2b <- group(xset2, minfrac = 0.50, bw = 5)
xset3 <- fillPeaks(xset2b)
reporttabACP <- diffreport(xset3, "A", "B", "groupdensity7550", 1000)

group.mzClust

Code: [Select]

xset <- xcmsSet(method="matchedFilter", fwhm=10, snthresh=3, nSlaves=4)
xsetb <- group(xset, method="mzClust", mzppm=25)
xset2 <- retcor(xsetb, family="s", plottype="m")
xset2b <- group(xset2, method="mzClust", mzppm=25)
xset3 <- fillPeaks(xset2b)
reporttabACP <- diffreport(xset3, "A", "B", "groupmzClust", 1000)

thanks for your help.

Re: problem with mzClust

Reply #1 – September 19, 2012, 07:19:46 AM

Hi,

mzClust is designed for alignment on single spectra samples stored in
a given xcmsSet, i.e. direct infusion data.

It implements the algorithm described in

Saira A. Kazmi, Samiran Ghosh, Dong-Guk Shin, Dennis W. Hill and
David F. Grant
_Alignment of high resolution mass spectra: development of a
heuristic approach for metabolomics_.
Metabolomics, Vol. 2, No. 2, 75-83 (2006)

and completely ignores retention time.

Yours,
Steffen

Re: problem with mzClust

Reply #2 – September 19, 2012, 08:41:19 AM

Hi,

only in addition to Steffens post, I would recommend to use centWave for peak detection,
because it works much better on QTOF data.
So change to:

Code: [Select]

xset <- xcmsSet(method="centWave", snthresh=3, nSlaves=4)

So long,
Carsten

Re: problem with mzClust

Reply #3 – September 20, 2012, 01:49:50 AM

Thanks for yours replies and your help and sorry for the error I did about mzClust.
Concerning the centwave method, we are developping the analysis method and will try it later but I effectively saw that it is
better for HRMS and QTOF data.

I have another questions about xcms grouping and RT correction. I've just seen that a "new" grouping method was set up named nearest. Is it adapted to LC-MS data and is anyone has already tried it ?
And for RT correction, I hesitate between loess and obiwarp methods. Knowing that my LC system causes high retention time deviations, which method would you recommend ?

Many Thanks

Ben