Skip to main content ↑ ↓

Messages

This section allows you to view all Messages made by this member. Note that you can only see Messages made in areas you currently have access to.

Messages - Jan Stanstrup

121
R / Re: DOUBT ABOUT MS^E
Msconvert from Proteowizard should be able to do that. Make sure to use the lockmassRefiner filter to get calibrated data.
122
XCMS / Re: How can I do the processing data? normalization?
Not familiar with XCMS online but loess regression in XCMS I only know from the retention time alignment while fold change would be for the stats.
You'll have to be a bit more specific to get help is my feeling...
124
XCMS / Re: How can I do the processing data? normalization?
1) Convert with proteowizard's msconvert to mzML
2) Use the XCMS package in R to pre-process data. Probably use CAMERA too for annotation.
3) Use whatever normalization and stats in R is appropriate for your data.
127
RMassBank / parser for massbank record --> R (list)
RMassBank first creates a list structure for the info that goes in the RMassBank record, but is there a way to go the other way?
--> A parser to read the RMassBank txt files into an R friendly structure.
128
XCMS / Re: peak shape/symmetry?
I don't think there is an easy way as you suggested yourself.
You can do what you suggest yourself doing something like this:


Code: [Select]
library(faahKO)
library(xcms)



cdfpath <- file.path(find.package("faahKO"), "cdf")
files <- list.files(cdfpath, recursive = TRUE, full.names = TRUE)

xraw <- xcmsRaw(files)

xset <- xcmsSet(files, verbose.columns=TRUE, method="centWave")
xset_g <- group(xset)


idx <- groupidx(xset_g)


out <- list()

for( i in seq_along(idx)){
  peak_select <- peaks(xset_g)[idx[[i]], ]

  # the max peak for all samples. Otherwise find max "sample"-wise
  out[[i]] <- peak_select[which.max(peak_select[,"into"]), c("sample","egauss","mu","sigma","h")]
}

129
XCMS / Re: From xcmsRaw to xcmsSet
The normal pipeline is to use xcmsSet that just takes file paths as input.
An xcmsSet can be subsetted using split and merged with c. Peak grouping info will be lost.
You'd normally use findPeaks to get peaks from a single xcmsRaw object but there seem to be no way to go from xcmsPeaks (output of findPeaks) to xcmsSet object. At least that I know of. It could probably be extracted from the source code of xcmsSet though.
133
XCMS Online / Re: Analysis peak intensity table after XCMS
You are talking about fragments/adducts? In that case you are looking for the R package CAMERA. It is meant to find which peaks belong to the same compound and annotate their relationship if possible. From that you could then filter your list if you wanted.
134
Publication Committee / Re: New Metabolomics Society Journal - Your Input Needed!
Quote from: schymane – on
Just a quick comment on Jan's wishlist - why double blind review? Why not consider the other way and have open review - to get back to #transparency :-) As a reviewer I appreciate knowing who the work comes from, it is often obvious (but not always and guesswork is not a great idea). If I want to discuss the work with authors beyond the review, I would like the opportunity to do so; I have also considered identifying myself in some reviews, my husband always does and receives very positive feedback. I feel this way is more constructive than double blind, but that's my personal opinion.

Because I think it is important to avoid bias also in the review process.
Of course I cannot know if there is a real problem with that but I am personally convinced that high profile researchers often get away with things other researchers might not (and that they shouldn't); lack of openness about methods, "trust me" kind of replies to review points, borderline appropriate papers for the journal etc. At least I have seen what I considered such kind of biases.

There might also be a bias from the reviewer being more likely to accept unclear methods and speculative conclusions if they already trust the authors. Also gender bias could be an issue.

Metabolomics is a small community so I would find it prudent to avoid favoritism, unconscious or not. This could also make it easier to find reviewers without conflicts in a narrow field.

I can see that generally people are moving to open reviews but it is my personal gut feeling, without much evidence, that it does not lead to better reviews. With open reviews you might tend to make more thorough reviews, but perhaps less likely to shoot down papers that really is not worth the time of a thorough review in the state it was submitted.

I am not sure "positive feedback" to reviews is really a metric worth aiming for...
I do accept that I could be completely wrong about this idea.


I just found this old paper on the subject that suggest there is some effect of blind reviews: http://www.unifr.ch/wipol/assets/files/PhD%20Course/Blank1991.pdf
135
XCMS / Re: Peak filling - an example of strange results
Sorry for the late reply.

The only parameter that seems a bit off to me is the ppm. Even with orbitrap that is very low. Keep in mind that all scans, even the ends of a peak, need to be within this. 10 or even 20ppm might be better.
You might also consider setting profparam but I don't think that could be your problem here (more about that here: http://www.metabolomics-forum.com/index.php?topic=598.msg1853#msg1853).

Quote from: chandlerjd – on
I really don't see anything I would deem a peak. Is Xcalibur adequate to look for this?
Yes. Are you looking at the mass spectra? Either there are peaks there or there is not. They are something like <5 % of the main peak if they are there so you have to zoom heavily.


Quote from: chandlerjd – on
Why is Orbi data prone to the shoulder peaks compared to other types of MS?
Because it is an artifact of the Fourier transformation as far as I have understood.
I have not found a paper that describes the problem in relation to orbitrap so I am not 100 % sure about the origin of the peaks. But my understanding/guess is the following:
I believe the problem is the phenomenon know as ringing (http://en.wikipedia.org/wiki/Ringing_artifacts). Other people call them ripples (I think they are discussion the same thing here http://www.shimadzu.com/an/ftir/support/tips/letter15/apodization.html). I have also seen them called simply orbitrap artifacts.
It is only visible in orbitrap data for very large peaks as small random mass peaks ~0.01 Da from the main peak.
Since these artifacts don't usually appear nicely symmetric as you would think from the pictures in the links above I assume many of them fall under some lowpass threshold during centroiding. But I guess it could also be that they are not symmetric in a complex signal. I don't know.



About the sorting: Are they more or less sorted by mz? I think that the original peak picking is sorted by mz, but not sure what happens when grouping.