What type of data should be used?

In general, I would recommend to work on clean data, but it is not an absolute requirement. The main things of importance are:

if you use a clustering step, check that the clustering is at a high enough level that it does not get impacted by batch effects (e.g. with the testCV function)
the data for which quantiles get aligned (i.e. every cluster, or the whole dataset if no clustering is applied) should be reasonably similar in percentage positive for all markers you want to use (so that's why we typically use a control sample)

In my experience clustering gives better results on cleaned data, and doublets and dead cells might also differ a bit more between the control samples, resulting in different distributions.

On the other hand, sometimes you want to normalize first to ensure you can have a more robust preprocessing pipeline for cleaning the data. So as long as these main requirements are quite ok, you can also apply it on raw data.

It is important data is transformed for correct clustering and quantile alignment. Either you need to pass a transformList, or you can also transform upfront, save these preprocessed fcs files and set the transformList to NULL.

All the best, Sofie

On Thu, 19 May 2022, 10:45 mattvtcea, @.***> wrote:

Hi, I would like to apply CytoNorm to my mass-cytometry dataset but I would like to know more about the data used. Do you use raw FCS files (after beads normalization and debarcoding) or do you "clean" your data first (singlets, live cells etc...)?

Best

— Reply to this email directly, view it on GitHub https://github.com/saeyslab/CytoNorm/issues/32, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAOS725LUQGIRYIEUL3QEDLVKX5SFANCNFSM5WLNYILQ . You are receiving this because you are subscribed to this thread.Message ID: @.***>

saeyslab / CytoNorm

What type of data should be used? #32