Batch Correction loess removes blank samples from dataMatrix

When I tried to use Quality metrics after a batch correction, I do have an error link to some samples that are missing in dataMatrix. After investigation all missing samples have sample Type set to blank that were removed by the batch correction function.

Can someone explain why blanks are removed?

Well, they are removed because the linear/lowess/loess methods do not currently handle samples other than "pool" and "sample". You currently have two options:

  • retag your blanks as "sample" for the Batch correction step
  • use the all_loess methods that handle blanks.

Please note that handling blanks in linear/lowess/loess methods is on my list of ToDo along with other improvements about these methods.

1 « J'aime »

Hi Melanie,

Thanks for the answer. Finding a "good" blank management is an important question/issue in metabolomics.

Hi @yguitton

Just a short notice to tell you that the new version of the tool (3.0.0), available on, now handles blanks in the linear/lowess/loess methods:

  • if you want them to be corrected but not to contribute to the normalisation in any other way, you can tag them as blank samples in the new 'Sample metadata file coding parameters'
  • if you want them to be considered as any non-pool sample, you can also choose to add them as a possible tag for samples (in the same new section).

See you on :wink:

1 « J'aime »