search

Functional-Metabolomics-Lab / FBMN-STATS

FBMN-STATS: A hitchhiker's guide to statistical analysis of Feature-based Molecular Networks
BSD 3-Clause "New" or "Revised" License
32 stars 25 forks source link

transform before plotting PCoA/PCA? #137

Open marwa38 opened 3 months ago

marwa38 commented 3 months ago

Hi

Could you please clarify if you transform your samples before plotting PCA/PCoA?

Thanks Marwa

abzer005 commented 3 months ago

If you're asking about preprocessing steps like log transformation, scaling, or other standardization methods, we are not applying any of those for the PCA or PCoA.

The data is used in its current state, so whatever cleanup or preprocessing (e.g., normalization or scaling) has been done prior will be reflected in the PCA. For example, if you choose to perform PCA on raw data, the data will not be scaled.

Hope this helps.

marwa38 commented 3 months ago

Will check if I can do log transformation beforehand in the GNPS, as I use directly the output of GNPS using task id for ananlysis as FBmnSTATS was mainly designed for. I do the normalisation step in your tool later already but for visualisation of PCA or PCoA it depends on the dataset if it is parametric or not. Transformation is a way to help the dataset that not meeting the parametric assumptions to be visualised in a better way.