Batch Correction changes all values to 0

[misch91]

Hey,
I am working on a dataset that was acquired in two steps and requires batch correction (see plot below).

Batch Correction Assessment report however suggests either changing all intensities to zero or touching only certain samples (see plots below). This happens regardless of the chosen window (I tried everything between 2-30). I first thought the preview must be mistaken but it turned out that running batch correction here really delivers the previewed correction.
There were no errors before, data had been imported and matched with metadata correctly. I suppose it is some bug within the batch and run order module?

[Gscorreia89]

Hi,

Interesting, definitely its misbehaving - Would it be possible to get a small example set to replicate this?

Thanks,
Goncalo

[misch91]

Dear Goncalo Here you can find the files and the script that led to the problem: https://drive.google.com/drive/folders/1a1AZti0J0xgxhAeJ93dzv35Io-ask0Me?usp=sharing It is a MS-DIAL processed output and for this I used the import script that I created in the branch msImportUpdates. The import however works the way it should, the data table and intensities have been imported correctly. I have had the same problem when importing as CSV feature table. Thanks for taking a look into it and best wishes Michael

[Gscorreia89]

Hi,

Found out the problem and a fix. The code is infering (correctly) the type of the data matrix as int and the loess correction is no longer modifying it to float (possibly due to some change in numpy/statsmodels?), which causes rounding to 0 whenever a value is below 1 - any value below the SR basically. The data matrix type (.intensityData) should be always float to avoid this issue.

For now, it can be patched by running msData.intensityData = msData.intensityData.astype(float) immediately after importing the dataset, but I will add a permanent fix.

[Gscorreia89]

Fixed by 005047b