mzAlign vs PeakAlign

[Ethan Yang]

Hi All,

I am having some difficulties figuring out the difference between mzAlign and PeakAlign. I have always used PeakAlign after PeakPick as suggested in the Workflow, so where/when is it more appropriate to use mzAlign?

Any input/guidance would be highly appreciated!

Sincerely,

Ethan Yang

[foellm...@gmail.com]

Hi Ethan,

mzAlign aligns spectra in profile mode to reduce mass shifts between them and should therefore be applied before peak picking. PeakAlign is required after PeakPicking but might not change much if mzAlign was already applied. For my low mass resolution peptide imaging data I perform  the three steps directly in this order (with more steps before and afterwards though): mzAlign - PeakPicking - PeakAlign

HTH

Melanie

[Ethan Yang]

Hi Melanie,

Thanks for sharing your workflow. It makes much more sense to me now, and I would love to make use of this function as a post acquisition internal calibration step. However, after looking through the thread and other documentation, it seems that this function is currently still only able to take in reference spectrum, so we will have to create a synthetic spectrum for this function. Is this your approach as well?

Sincerely,

[foellm...@gmail.com]

True there is also re-calibration to bring the peaks into a correct position on the m/z axis. In the latest Cardinal version (2.8) it is possible to just provide the known m/z values for re-calibration instead of a synthetic spectrum. 

So I first align the shits between the spectra and then bring the peaks to the right m/z position in a second step. Unfortunately, currently these provided m/z values will stay a sort of a "reference", therefore after peak picking it's important to create an average spectrum which is then used as reference during peakAlignment.

Here is some exemplary code:

## alignment

TMAs_mzaligned <- mzAlign(TMAs_s_b, tolerance = 200, units = "ppm")

TMAs_mzaligned <- process(TMAs_mzaligned)

## mz recalibration

calibrant_mz <- c(805.42, 1296.69, 1347.72, 1570.68)

TMAs_recalibrated <- mzAlign(TMAs_mzaligned, ref=calibrant_mz, tolerance=200, units="ppm")

TMAs_recalibrated <- process(TMAs_recalibrated)

## peak picking, alignment, filtering

TMAs_picked <- peakPick(TMAs_recalibrated, method = "simple", SNR=5, window = 10, blocks = 500)

TMAs_picked <- process(TMAs_picked)

mean_MSI <- summarizeFeatures(TMAs_picked, FUN="mean")

fData(TMAs_picked)$mean_spectrum <- as.numeric(spectra(mean_MSI))

TMAs_pp <- peakAlign(TMAs_picked, ref="mean_spectrum", tolerance = 200, units = "ppm")

TMAs_pp <- peakFilter(TMAs_pp, freq.min = 0.01)

TMAs_pp <- process(TMAs_pp)

[Ethan Yang]

Hi Melanie,

Thanks for the example code! I noticed that you have provided 4 peaks. Is there a recommended amount? What kind of algorithm is used in this process? Is it something similar to the quadratic/cubic realignment when we do initial calibration before data acquisition?

Sincerely,

[educh...@gmail.com]

Hi Ethan,

the most reference peaks the better. You need to cover the entire range with them, ie, if you have a scan between 700-2000, distribute 4-5 peaks more or less uniformly over this range (750, 1100, 1500, 1900).

I've a question to both of you, Melanie and Ethan: when i perform my recalibration with Melanie code, RStudio session just crashed and doesn't work. Did you tried it? I've Cardinal2.6.0 installed and 256GB RAM available

This is my code:

#refpeaks
calibrant_mz <- c(740.40, 842.51, 1046.54, 1570.68)

#Alignm
tissue_mzAlign<-mzAlign(tissue_pre, tolerance = 2000, units = "ppm")
tissue _mzAlign<-process(tissue _mzAlign)      ######HERE RSTUDIO CRASH#####

#Recal
tissue _recal<-mzAlign(tissue _mzAlign, ref=calibrant_mz,

                                      tolerance = 200, units = "ppm")

tissue _recal<- process(tissue _recal)

#picking
tissue _picked <- peakPick(tissue _recal, method = "simple", SNR=5, window = 10, blocks = 500)
tissue_picked <- process(tissue _picked)

#peakAlign
mean_MSI <- summarizeFeatures(tissue _picked, FUN="mean")
fData(tissue _picked)$mean_spectrum <- as.numeric(spectra(mean_MSI))

Best regards!

[kbemis]

As Edu says, the more reference peaks, the better. There is no recommended amount. One would have to perform an experiment with known calibrated data and references and compare the results.

This is an experimental, unpublished method, although the implementation is fairly standard. It uses local regression to estimate the mass shifts, and then cubic spline interpolation to perform the spectral shifts.

-Kylie

[Melanie Föll]

Hi Edu,

my mzAlign code will only work in Cardinal v2.8. For v2.6 you need to generate a reference spectrum of your calibrants as described in another conversation here.

Best,

Melanie

[ismaelu...@gmail.com]

Hi Melanie,

I'm working with Edu in the mzalign code but we have a problem running the code, I'm working with 2.8v of Cardinal, when we run the code (below) the R session just stopped.

I let you one pic of the error.

## CODE ##

decalcified_Cut <- readMSIData('pulmon155_1.imzML', type='MSImagingExperiment')

decalcified_Cut_pre<-normalize(decalcified_Cut, method='rms') #metodos disponibles: tic, rms, reference

decalcified_Cut_pre<- process(decalcified_Cut_pre)

calibrant_mz <- c(740.40, 842.51, 1046.54, 1570.68)

decalcified_Cut_mzAlign<-mzAlign(decalcified_Cut_pre, ref = calibrant_mz ,tolerance = 200, units = "ppm")

decalcified_Cut_mzAligned<-process(decalcified_Cut_mzAlign)

decalcified_Cut_recal<-mzAlign(decalcified_Cut_mzAligned, ref=calibrant_mz, 

                                      tolerance = 200, units = "ppm")

decalcified_Cut_recal<- process(decalcified_Cut_recal)

Thanks for your attention.

- Ismael

[Melanie Föll]

Hi Ismael,

your code looks fine to me but unfortunately the error doesn't give any information. Maybe you were running out of memory. You could crop your dataset to a few (hundred) spectra and test the pre-processing pipeline with the small dataset before you run it on the complete data.

Best,

Melanie

[ismaelu...@gmail.com]

So much thank you Melani, that was our problem.

When we proved it my computer don't have the enought memory and Edu had a previous version of Cardinal and made conflict but we didn't know, so we thought that the problem was another thing.

Thanks once again.

Best,

Ismael