Compare two more datasets

[Ivo Kwee]

[from Shahab]

Hi there,

Just wondering what would be the best way to compare two (or even more) TMT proteomics data sets, from independent runs? Should we normalize the intensities beforehand? Can we use batch conversion for less bias?

Thanks,

--

Shahab Mirshahvaladi,

[Ivo Kwee]

There are different ways you can do it

1. Merging the data sets, then assess for batch effects and if necessary correct for them. This method is if you want to see the heatmap or clustering of all samples together. The disadvantage is that you must be really careful about the batch effects and the correction will "alter" your data. Another "disadvantage" is that you need to recalculate all statistics. If you want to be conservative, create the contrasts only within the batches, otherwise if you trust the batch correction, you can create contrasts by mixing samples from different batches. Intensities don't have to be normalized by hand, because the Playground is doing sample-wise normalization by defaults (i.e. CPM+quantile).

2. A more conservative, safer approach would be to only compare the data sets only in their fold-change. This would avoid dealing with batch effects because the data sets are analyzed seperately. The disadvantage is that you don't have an integrated heatmap with combined samples. Advantage is that you don't need batch correction. You can do this with the "compare datasets" feature (first turn on the beta features in the settings).

Ivo