I have two questions about using msqRob2 for differential protein abundance estimation (using output from MaxQuant):

1) Is there a reason that msqrpob2 uses raw Intensitities, rather than the normalized "LFQ Intensities" from MaxQuant?

2) I would like to confirm the correct model design that includes blocking. I was using this tutorial as an example: https://statomics.github.io/msqrob2Examples/mouseRCB2.html and saw that the model was specified as "formula = ~ celltype + mouse", in which the factor of interest (cell type) is listed first and the blocking factor (mouse) is second. However, in every other linear model I've used (e.g., limma, deseq2), the blocking factor is listed first, then the factor of interest (i.e., ~ mouse + celltype). Which is the correct approach?

Thank you!

Dear Jane,

1) msqrob2 can start from LFQ intensities, but our robust summarisation when starting from peptide intensities outcompetes LFQ summarisation on benchmark datasets. (e.g. see the wrap-up in my course on proteomics data analysis, https://statomics.github.io/PDA/pda_robustSummarisation_peptideModels.html)

2) For statistical inference and fold change estimation, it does not matter if you specify the blocking factor first or last. The resulting parameter estimates and standard errors will be equivalent.

Best

Lieven