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 - krista

17
XCMS / Re: centwave: prefilter = c(K,I) or noise
Hi Cole,

OK. The distinction makes sense to me, although I still cannot envision a case where two different values would be useful. Thanks for the clarification.

Krista
18
XCMS / centwave: prefilter = c(K,I) or noise
Hello,

In the documentation on findPeaks.centWave  there seem to be two different parameters to set an intensity threshold during the ROI detection step.

1) prefilter = c(K,I) which keeps mass traces if they contain at least K peaks with intensity >= I.
2) noise = which is described as an optional parameter to remove centroids with intensity < noise

I understand both of these ideas. What I don't understand is why there are two ways to set the lower intensity bounds during the ROI detection step. Is there a case where the intensity value in the prefilter would be a different number than the noise parameter? Or, have I misunderstood the two parameters?

Thanks,
Krista
19
metaXCMS / alignment after Common Features
Hello,

I am curious about the order of steps in metaXCMS. As I understand it, the 'find common features' occurs before the alignment/retention time correction step. I am a little confused by this. To me, it seems like you would want to do the alignment & retention time correction across the different datasets first, and then find the common features. Am I missing something here?

Thanks,
Krista
20
metaXCMS / Re: Alignment + EICs
I can also add that I was able to get rid of this error by starting all my analyses from one common upper level folder. My first attempt at leaving all the XCMS files in place, but having them in different directories also resulted in the error about directory level not being flat.
21
XCMS Online / Cloud plot: upregulated vs. downregulated
Hello,

I have just discovered that the Cloud Plots in XCMS Online are a fantastic way to explore features of interest in a dataset. However, the terminology of 'upregulated' and 'downregulated' is potentially a bit confusing. Clearly I have uploaded two datasets which are entered into the online tool as 'Dataset 1' and 'Dataset 2'. Based on the box and whisker plots, 'upregulated' seems to be higher in dataset 2. However, I could not find any documentation actually stating this. I had actually expected upregulated to be higher in dataset 1.

I wonder if the phrasing in the Cloud Plot could be more broadly described in the figures as "higher peak heights in dataset 1 (or 2)"'? My reasoning is as follows. For an experiment in the lab, I the datasets are presumably  (1) one set of experimentally altered samples compared to (2) an un-manipulated control. However, for a set of field samples, the description does not always fit into experiment vs. control. For example, I could have fish from Lake A compared to fish from Lake B. Neither one is a control per se, but there are differences across each lake.

This is not really a question per se since I mostly wanted to suggested a slightly different phrasing for the cloud plots. However, if I am wrong about 'upregulated' meaning that peak heights are higher in dataset 2 I would like to know.

Thanks,
Krista
22
Other / compare metabolites across experiments
I am hoping to open a broad discussion on the best way to analyze metabolites across experiments. In particular, I am thinking about such comparisons in the context of storing untargeted metabolomics data in a database. My thinking is as follows:

1. Unidentified compounds (i.e. compounds that are not in existing databases) may be present in multiple experiments. In this context, an 'experiment' might be samples collected in the lab or from a set of field sites.
2. The presence/absence of these unidentified compounds across experiments will determine which compounds will be most interesting to identify. For example, a compound at mass X is found in all samples of type A. The mass X is not found in any databases of known compounds, but the frequency that the compound is found means that it is one that I should try to identify.

However, there are some major caveats with this since you have to decide how to align features (both m/z values and retention times) across sample runs and potentially even instruments. This is a major issue. But if we acknowledge the caveats, what is the best way to align metabolites from different experiments in a manner that allows us to store unknown metabolites in a database?

Alternatively, is it better to populate the data base with the m/z values and retention times from each experiment, and then setup a search engine to look within a given error window for m/z values and retention times?

Krista
23
CAMERA / Re: groupCorr, calcIso = TRUE
The changes to CAMERA version 1.9.7 now allow me to process my data without any errors.

Thanks for the help & new program version,

Krista
25
CAMERA / Re: groupCorr, calcIso = TRUE
Hi Carsten,

Thanks for looking into this.

I could easily be missing something here, but when I look through BioConductor, I only find CAMERA version 1.9.6 - and I did check that, the error still appears with that version.
 
Krista
26
CAMERA / groupCorr, calcIso = TRUE
Hello,

I am trying to use CAMERA with LC-MS (FT-ICR) data, and have an interesting problem. Oddly, this problem is only appearing in one ionization mode, so it is apparently a data dependent problem. However, I can't figure out where the problem begins. The actual problem comes while I am at the groupCorr step.

By this point, I have done:
Code: [Select]
xgN<-group.nearest(xs)
xgF <-fillPeaks(xgN,method = "chrom")
xsa <-xsAnnotate(xgF)
xsaF <-groupFWHM(xsa)
xsaFI <-findIsotopes(xsaF)
xsaC <-groupCorr(xsaFI,cor_eic_th=0.75,pval=0.05, graphMethod="hcs",
  calcIso = TRUE, calcCiS = TRUE, calcCaS = FALSE)

I then get an error after  'calculating isotope assignments in 290 groups'. The error is:
 % finished: Error in rbind(resMat, cbind(x = tmp2[, 1], y = tmp2[, 2], cor = 1, ps = i)) :
  number of columns of matrices must match (see arg 2)

traceback() says:

 rbind(resMat, cbind(x = tmp2[, 1], y = tmp2[, 2], cor = 1, ps = i))
8: calcIsotopes(object)
7: calcIsotopes(object)
6: .local(object, cor_eic_th, pval, graphMethod, calcIso, calcCiS,
      calcCaS, psg_list, ...)
5: groupCorr(xsaFI, cor_eic_th = 0.75, pval = 0.05, graphMethod = "hcs",
      calcIso = TRUE, calcCiS = TRUE, calcCaS = FALSE)
4: groupCorr(xsaFI, cor_eic_th = 0.75, pval = 0.05, graphMethod = "hcs",
      calcIso = TRUE, calcCiS = TRUE, calcCaS = FALSE)
3: eval.with.vis(expr, envir, enclos)
2: eval.with.vis(ei, envir)
1: source("Tps4c_pos.r")

If I set calcIso=FALSE, then the analysis proceeds without error.

I have also tried going back a step and running groupCorr on the data without the isotope information, and then it notes there is not isotope/isotope annotation (which makes sense), but the analysis goes to completion:
Code: [Select]
xsaF <-groupFWHM(xsa)
xsaC <-groupCorr(xsaF,cor_eic_th=0.75,pval=0.05, graphMethod="hcs", calcIso = TRUE, calcCiS = TRUE, calcCaS = FALSE)

There is apparently something happening at the findIsotope step, but I have been trying different parameters with no luck. I suppose I could just use the calcIso=FALSE results, but since I was able to get this to work in negative ion mode, I am perplexed.

I am running:
CAMERA 1.8.2
xcms 1.26.1
R 2.13.1

Any thoughts and/or places where I should be looking?

Thanks in advance,
Krista
28
CAMERA / findIsotopes and ppm
Hello,

I am trying to use the findIsotopes function within CAMERA. I am pretty excited to use this function as it will certainly help decide which features I will focus on for my downstream analysis.

However, I am wondering about the implementation of the 'ppm' variable. Following the syntax on the help page, I am doing the following:

xsaFI <-findIsotopes(xsaF,ppm=1)

The first set of isotopes are:
293.1617 [1][M]+
294.1755 [1][M+1]+

I interpret this as the 294.1755 is the 13C match to the 12C compound that is 293.1617.

However, if you change one of the carbons in the 12C compound to be 13C, the calculated mass would be 294.1650. By my math, the difference (in ppm) between 294.1650 and 294.1755 is 35.7 ppm, which is way more than the 'ppm=1' I requested.

The question is, is the error actually defined somewhere else in the processing? Or, what am I missing?

Thanks,
Krista
29
XCMS / Re: centWave & ROIs
This worked:

pdf(file="someFile.pdf")
xs<-xcmsSet( ... )
dev.off()

It seemed to ignore par(mfrow=c10,5)), but that was fine because I end up with a PDF with one page per feature.

Thanks!

Krista
30
XCMS / Re: centWave & ROIs
Is there a way to save the individual plots? I have been using the manual ROI finding with great success, but now I would like to go one step further and see the results from each wavelet/peak finding step.

The sleep value is helpful. However, there is a tradeoff between wading through lots of figures with a high sleep value, or having them go by too fast in order to get through all of the figures.

In other words, based on some of the downstream analysis, there are specific m/z values and retention times I want to check. However, while I did eventually figure out that I could use scanrange = cbind(#,#) [and not c(#,#)] with findPeaks.centWave, I still have lots of figures to wade through.