Possibility of integrating multiple markers

[ZYX0711]

Hi,
Thank you for the packages! I have a question about applying Infinity Flow to a slightly different experimental design.
I have 3 independent flow panels that share a backbone set of markers, but each panel has multiple additional markers:
Panel1: backbone + Marker1+Marker2
Panel2: backbone + Marker3
Panel3: backbone + Marker4
Infinity Flow is designed for backbone + 1 “infinity” marker per panel, but my panels have 1–2 extra markers per panel, not strictly one.
I’d like to know:
Is it acceptable practice to run Infinity Flow sequentially per marker (i.e., one model per marker) and then merge the predicted markers into a single high-dimensional dataset for UMAP/clustering?
Do you foresee any limitations or caveats when doing this and are there any recommendations for best practices?

Best,

[ebecht]

Hello,

The package is designed for single marker imputation. In theory what you propose is sound but I'm not sure it'll be straightforward to perform using the infinityFlow package. We sample distinct random cells for training and predictions, so you'd have to make sure the same cells are sampled. I think that if you set a similar seed before each run that should work but am not 100% sure.

Best
Etienne

[ZYX0711]

Hi,
Thanks for the suggestions! I will try it out.

Best,