Hi David,

I think, I see the problem. You are using XCalibur software (Thermo)? And you performed MS and MS/MS-scans during the LC run. Do the following:

1. Open the raw-file with this software. 2. Click in the upper right corner of the "LC-data" (upper cell) to choose this line. 3. Click with the right mouse buttom in this cell and choose "Ranges". 4. Now choose in the scan filter ONLY the MS (not MS/MS etc.) line (most of the time it is the first or the last line). 5. Click "OK". 6. Now you can average (as you probably already tried) by choosing the lower cell ("MS-data"; click in the upper right corner) and drawing a line in the upper cell with the left mouse buttom in the requiered time range.

I hope, this helped.

I suppose it depends on the software and vendor of your equipment. With Waters MassLynx you can easily right click and drag across a peak in a chromatogram of multiple ions and it will give you an average spectra of all the ions in that peak.

Which software are you using? Most of the software allow you to select a range to give average spectrum. Check your user manual.

Hi David,

I think, I see the problem. You are using XCalibur software (Thermo)? And you performed MS and MS/MS-scans during the LC run. Do the following:

1. Open the raw-file with this software. 2. Click in the upper right corner of the "LC-data" (upper cell) to choose this line. 3. Click with the right mouse buttom in this cell and choose "Ranges". 4. Now choose in the scan filter ONLY the MS (not MS/MS etc.) line (most of the time it is the first or the last line). 5. Click "OK". 6. Now you can average (as you probably already tried) by choosing the lower cell ("MS-data"; click in the upper right corner) and drawing a line in the upper cell with the left mouse buttom in the requiered time range.

I hope, this helped.

what I forgot: the "MS"-line in the SCAN FILTER might be: "FT + p NSI Full ms [300.00-1800.00]"

Did it work?

Check the software manual and also follow the suggestions of Dr. Gunnar Schwartz.

I've actually already done what you (Gunnar Schwarz and others) suggested (the image I've provided shows the result of these steps). I've realized that my question was not clear enough.

I know how to get an averaged spectrum using the commercial software "Xcalibur" but I would like to "mimic" / reproduce the averaging behaviour with my own code/program. Since I can't average over thousands of peaks and hundreds of measurements, I'd like to automate it, but am missing the proper averaging function. Should it be a polynomial, exponential, etc. function? How to bin m/z values? Meaning, that an m/z value from the one scan will slightly deviate in another even if it belongs to the same isotopic species. I hope it is now clear what I mean.

Thank you for your help!

Otherwise you can export data (numeric values) of the chromatogram as CSV files and perform the averaging in EXCEL.

Assuming you don't have within your software the metabolomics tools, for aligning the chromatograms you can use free software like MetAlign or XCMS software from METLIN (Scripps) page. You can find links for other free softwares in Metabolomics Society page.

Regards

ad Olimpio Montero:

well yes of course. But now my question remains, HOW should the data be averaged?

A simple sum of all values divided by the number of values does not produce a satisfactory result. Thus a weighted averaging function need to be employed. I would like to know which weighted averaging function to use.

I unknow the exact answer to the last question. I think it would be the best that you use some free software I indicated. Nonetheless, I think the best procedure, for example to be carried out in EXCEL, should be to normalize all the values to the sum of the whole set values. Indeed, those software work in such a way although I don't know whether the values are previously weigthed with a given function.

By using the indicated softwares you get aligned chromatograms with the mass spectrum corresponding to every time point average already. Then you can get average the whole set of chromatograms after CSV export.

pymzML offers such a functionality.

For more information see http://pymzml.github.io/usage.html#advanced-usage

Hi David,,,try with the ACD/labs program....

All: you do not quite understand what David is asking.

We (naively) assume that the m/z axis is always the same in every scan, but it is NOT: a given spectrum may be missing points at the termini (low and high m/z), it may not have perfectly identical spacing between m/z point, and the "offset" of the m/z sampling interval can change.

For example, one spectrum will have a discrete m/z-intensity data point at 118.006123, but 3 seconds later that data point will be at 118.00724. See here (actual data) for many sequential scans, overlaid: http://oi42.tinypic.com/x10fvq.jpg

I've talked with the software people from large MS companies, and they are well aware of this issue but they didn't want to divulge their processing techniques.

David: my testing has shown that true peak motion on the m/z axis is less variable than the m/z sampling drift. This implies that it's probably not a good idea to bin with anything less than very small intervals, and those intervals may need to be intelligently selected according to a time series (see above image).

If you can spare the CPU time, you get very nice results by converting the discrete m/z vector into a continuous spline/lowess and resampling however you want. I used approaches like this for a previous publication (google hilmer, physical signal modulation). Conversion to a continuous function works best when you have not artificially cropped off baseline with the digitizer, because sharp transitions between "background" and peaks cause problems with interpolation.

may be you need to change MS saved format in profile.

did you try MetAlign (http://www.wageningenur.nl/en/show/MetAlign-1.htm)?

Thank you to Jonathan!