Quote from Rasmus Bro on April 11, 2023, 8:21 pm

Hi

We have not tried that but it may very well work. I am guess you should still try to work on intervals - maybe start by looking at one minute of data and learn what that leads to. You probably want to select intervals so that the number of components stay below around ten.

But it sounds like something you can easily spend a PhD or two one!

Best of luck

Quote from Guest on April 13, 2023, 1:13 am

Thanks Rasmus,

Will try it out when we have a more complete set of the data.

Quote from Guest on April 26, 2023, 11:50 pm

In relation to using the fast GC-TOF data, how does PARADISe deal with the exact masses?  The TOF m/z data has 6-20 decimal places (when viewing the data in R it displays 5 decimal places) and there are some that if rounded to 1 decimal place would end up at the same mass, e.g. 89.02332, 89.04195. When trying to display an exact mass in PARADISe, even entering as, for example, 81.07 (rounded from 81.06989) PARADISe comments that it can't find that mass and selects for the next closest, 81.

Anyway, if you have any comment on how PARADISe deals with decimals in the mass ions that would be very helpful.

Thanks heaps,

Andrea

Quote from Jesper L Hinrich on April 28, 2023, 9:05 am

Hi Andrea,

The answer is pretty simple, PARADISe is build and tested on unit mass data, so importing from CDF will always round to nearest unit mass.

We have done some experimentation with non-unit mass, but some issues arise that are not present in the unit mass case. Two of the main issues are,

1. The data becomes extremely sparse (many zero values) and the computations have to exploit this to be feasible. PARADISe already does this to some degree, but working with sparse data arrays saved to disk is harder than it sounds.
2. A non obvious issue with high mass spectrometers is at some decimal point, they are no longer trust worth. For example, a fragment might be recorded as both 89.02333 and 89.02332 which then introduces a shift in the m/z axis. Handling such a shift elegantly and robustly is challenging. This issue becomes more present the more decimals you include.

Now, you can actually experiment a bit with non-unit masses, but this requires manually setting up your chromatography data as a .mat file. In this file, you need to ensure the m/z axis (variable mz) is given as non-unit mass and the data (variable Data) is of the correct size (it will be big). Then you can import this chromatography data into PARADISe, but the memory requirements and computation will increase significantly. We done some testing with a mass resolution of 0.1 and 0.01 which is manageable for experimentation, but it does require additional patience.

Best,
Jesper

Quote from Guest on May 1, 2023, 6:51 am

Thanks Jesper,

That makes sense, might be something we have a work through later.

Cheers,

Andrea