Hi.
I am working with flow cytometry data. I have two batches of PBMC samples with the exact same panel, 25 proteins.
I have a ML model for cell type prediction on one batch (=anchor) and trying to apply it to another batch (=query). They have slightly different coordinates, meaning distributions are shifted in the majority of channels. My idea is to use an anchor method and integrate the query into the anchor batch without changing its coordinates. This integration should change raw values of the query only.
Below, is the result I would like to see, but without changing anchor’s values (blue cohort):
I have tried FindIntegrationAnchors, CCA, from Seurat, it works not ideally, and it’s slow, i.e. anchor is 100k cells, query is 100k cells, 128 GB RAM was running for about an hour.
I tried to use MNNpy, but couldn’t install it on new macos (15.5). I will try it on a different machine soon.
Am I missing something? Are there other tools?
Thank you,
Emir.