Question About Batch Effect Correction in standR for GeoMx Data Analysis

[Rongtingting]

Hi!
Thank you for developing the package!
I’ve been following the tutorials provided in the standR documentation (Quick Start Guide and GeoMX Analysis Workflow), and I have a question regarding the batch effect correction step.
Based on my understanding of the tutorials, it seems that the spe_tmm object (after normalization) should be used in the batch effect correction step. However, in the examples provided, the tutorials use the spe object directly for this step.
Could you kindly clarify whether we should use spe_tmm or spe for batch effect correction? I want to ensure that I am applying the workflow correctly.
Thank you for your time and assistance! And looking forward to discussing more!

My understanding for batch effect correction step

spe_tmm <- findNCGs(spe_tmm, batch_name = "SlideName", top_n = 500)
spe_ruv <- geomxBatchCorrection(spe_tmm, factors = "biology", NCGs = metadata(spe)$NCGs, k = 5)

Tutorials show:

spe <- findNCGs(spe, batch_name = "SlideName", top_n = 500)
spe_ruv <- geomxBatchCorrection(spe, factors = "biology", NCGs = metadata(spe)$NCGs, k = 5)

[ningbioinfo]

Hi @Rongtingting , in my pass experiences, using it or not using it pretty much give similar outcome (DE genes). So it should work either way. If you want to test it, can for the tmm one, make sure have tmm norm.factor in the DGElist when doing DE, while for the non-TMM one, just leave norm.factor column as 1. But both of the linear model should include the W matrix from RUV calculation.

[Rongtingting]

Hi @ningbioinfo， Thank you so much for your answer!! I am not sure what you means in have tmm norm.factor in the DGElist when doing DE. Could you help to explain more on this step?

Bests,
Rongting

[ningbioinfo]

Hi @Rongtingting , so basically if you using y <- standR::spe2dge(your post-normalised spe) when starting the DE pipeline, the resulting y (a DGElist object) would have a "norm.factors" column inside y$samples. These will be the norm.factor linear model take into account when performing DE fit, in your case, would be TMM.