PARADISe for resolving untargeted GCMS data. See video guides on using the software here.


We have simplified the selection of components again. We now have enough data for developing automated models. Thanks to all of you who sent us their stuff.



It is advisable to first uninstall any prior version before installing a new one. Also note that you need to have the Visual C++ redistributable installed first. 


This version may hopefully be stable version. Please let us know if you run into trouble with this one.
This version fixes a bug in version 2.9 and 2.91 that made it impossible to import certain types of CDF files. Please let us know if you run into problems and we will try our best to fix it.


Download software for windows v. 2.91 (May 8, 2018).


Download software for windows v. 2.9 (Apr 27, 2018).

  • New orientation of elution profile plots to make it easier to see what is happening.
  • New colors indicating whether a profile is likely to be a peak or not. Color code added to the plot. These colors are not currently used but are just expressions of our work with the automated intelligent model we hope to develop in time.
  • Fixed handling of varying time points in different samples. Now times are interpolated to most frequent common time axis. Early and late parts are cut unless they appear in all samples.
  • Added the spectra of the pure components in the excel report.
  • Added auto-selection of intervals for the daring ones. Highly experimental and not so easy to use.
  • Added a better select-interval functionality. Now you can keep choosing without having to push the SelectInterval button all the time. Try and play with it; it really makes life simpler when you get the hang of it.

Download software for windows v. 2.6 (Feb 8, 2018).

  • There was a bug that stopped 2.3 from running at times. Should be fixed here.
  • Changed mass range imported from 1-1.000 to 1-10.000 (if possible)

    Changed interval waitbar to be a little more informative

    Data are now always saved in matlab format in the project folder. If you have to read in the data a second time, then reading the matlab file is usually easier.

    Changed the “show intervals” to be on as default now (every time you rezoom or similar, the “show interval” will take effect)

Download software for windows v. 2.3 (Jan 10, 2018).

  • Version similar to 2.2 but now it actually runs! Or at least is supposed to. 

Download software for windows v. 2.2 (Dec 21, 2018).

  • Added the option to say that a certain interval did not have coleution problems solved.
  • The elution profiles are colored a little funny in some plots. Bear with us. We are using a new feature to try and develop a tool to predict the niceness of an estimated peak. More to follow.

Download software for windows v. 2.0.0 (Dec 21, 2017).

  • Improved the import function. More varieties of CDF files should now be possible to import.
  • Added the ability to trim the retention time span to remove nonsense parts.
  • Added a button that will align the data. So far it is without any settings to adjust. If the alignment improves your data, it should be easier to select intervals. If not, you can click the button again and revert to the unaligned data. 
  • Parallel computing. Added parallel computing. If you have a multicore system, calculations should speed up now.

Download software for windows v. 1.1.9 (Nov 20, 2017). Changed button called 'Invalid'. Now it is called 'Converged' which is what the color really indicates. Added a new functionality for helping further development - see above.

Download software for windows v. 1.1.8 (Sep 10, 2017). Fixed problems with NaNs leading to missing plots and other

Download software for windows v. 1.1.7 (Sep 2017)




Please refer to the below paper when using the software:

L. G. Johnsen, P. B. Skou, B. Khakimov, and R. Bro. Gas chromatography – mass spectrometry data processing made easy. Journal of Chromatography, A 1503:57-64, 2017.


A little trick if you want to import your data through MATLAB: 

The GUI expects that the file loaded with the load .mat function contains the following:
FileNames - cell array with filenames
mz - vector with mz values
rt - vector with retention time (in minutes)
TIC - matrix with the TIC's (sample x scans)
M - 3-way array (sample x scans x mz)
BPC - matrix with basepeak chromatograms (sample x scans)




Frequently asked questions

Here are some of the questions we have had - plus some answers
Half peaks: Within an interval there might possibly be partial peaks not completely contained within the interval i.e. they are cut off at the beginning or end time points of the interval.  Are these partial peaks also included in the integration result for the main peak? 
Nope. Only the ones you select as chemical. The rest are sort of removed from the integration.
Baseline: How is baseline handled?
Well, ideally each source of information is removed. So baseline is removed in one or more components and so are tailing peaks. Each peak is integrated indirectly by the 'score' which measures the area under the curve. 
Baseline resolved
See the example above. To the left is the set of TICs with clear baseline as well as "half a peak" to the left. When a 3 component PARAFAC2 model is applied, the result is
Component 1 is the ‘target’ (ethanol) – light blue
Component 2 is a partial peak – dark blue
Component 3 is background noise (green)
Note that the target peak now starts at zero end ends at zero (practically).
In this case, only component 1 is tagged and the and the summed of the signal of that only will be the concentration. So no baseline and no neighbor peak.
Converged?: When choosing the best ‘Fit’ and ‘core consistency’ the green ‘converged’ button sometimes changes to a red ‘not converged’ sign.  What does it mean when it is ‘not converged’? 
The calculations have not ended optimally, so in essense you didn't get to the real solutions. You can try to recalculate the interval using more iterations. If you already allowed a fair number of iterations (say 2000 or more) then there is a chance that even though it says not converged, not much will change. Hence, you may go along with whatever solution you found if you are otherwise happy with it. For small or very complex data sets, it may be near impossible to get convergence.
Negativity: The weighted elution profiles often show negativity in the baseline, yet the TIC does not.  When the software is integrating the peaks, does it integrate the non-negative TIC? What is the meaning of the negative sections of the weighted elution profiles?
That's a complicated one. If you apply nonnegativity, then this should not happen, but for nerdy reasons, it can still be the case. Which then probably indicates a very complex problem. Trying to simplify e.g. by making the interval more narrow may help. There could also be other reasons so if you have a problem in this direction that is bothering you, then send it to us and we will check.
Redoing: Is it possible to go back in and edit the intervals set within the PARADISe analysis of a set of runs. It may be that a first attempt successfully integrates 75% of the compounds but others need further adjustment. At the moment, I cannot see how to do this without entering all the intervals again, provided, of course, that I have a manually written record of the intervals I used before. 
You can reload the intervals and then edit them. Observe the ‘Status’ column. When intervals are reloaded, status will be 2 (=’calculated’). If you change an interval, set the value to ‘1’ and the interval will be recalculated if you click ‘Start PARAFAC2’.
Integration: Will the 'flanks' be part of the integration?
Yes - in the dark green component below, the left and right most part would bias you concentration estimates