Patient variability

[Tregwiz]

Hi,

Thanks for the fantastic package and for all of your support with it.

I had more of a general question about catalyst. I am analysing a CyTOF patient dataset in which significant inter-patient variability is present. By this I mean that although two patients may exhibit a similar percentage positivity for a marker, the signal from the negative and positive populations for each patient varies significantly i.e. one patient may show a negative population up to 10^2 whereas the other patient shows a negative population up to 10^3. Bead normalisation has been correctly applied to this data and no batch effect is present.

My general question is, is Catalyst designed to account for such variability? Does it apply scaling by sample and assign the expression value based on the distribution of cells within this sample? Because I obviously don't want the same expression value to be assigned to a 10^3 cell for both patients, as this might represent a negative cell for one patient and a positive cell for another.

Thanks.

[SamGG]

CATALYST is based on the concept of pooling cells from different samples and analyzing the resulting set as homogeneous cells in term of marker intensity. So, there will be a merging of positive and negative cells in the example you described. This approach is widely used and seems to pose no problem in most of experiments.
Alterntively, you may consider to apply the clustering algorithm to each sample individually, which requires a meta-clustering in the end. Take a look at scaffold, grappolo, metacyto, etc. approaches.
I don't know which marker you are taking about, as I never saw cells having an intensity of 1000 and being negative in mass cytometry. I would have a look at batch effects in addition to patient variability.
Hope this helps.