Entering edit mode
11 days ago
MitchT
▴
10
Hi all, I am trying to tease out a rare cell type, ILC2s, from scRNA-seq data from human liver samples. Is it possible to modify an already existing signature matrix? Also having difficulty getting clean subclusters of NKs, ILC1, ILC2, ILC3s. Any advice would be greatly appreciated. I am using the most updated Seurat version for my workflow, with the end goal being a signature matrix for CIBERSORTx.
That's open-ended. Simplest explanation is that they are not present in the dataset. These rare subsets are often hard to capture unless your scRNA-seq used a dedicated FACS enrichment scheme, since the liver is mainly hepatocytes, LSECs, stellate cells and Kupffer cells. I would use canonical markers and just color the UMAP. See if the calles are there, and then decide whether the clustering is simple not capturing them, of if they're simply absent. If clustering is not working well, consider to maybe subset to lymphoid cells first, and then recluster, with a fresh and more focused set of variable genes, PCs etc.
Thank you. Sorry, I should have prefaced that I was able to find canonical markers of these cells in my main cluster, which is a mixture of healthy and diseased cells-combo of both CD45+ enriched and CD45- enriched. I subclustered from three clusters that showed strong markers for NK cells and filtered out any CD3, since NKs and ILCs do not express these markers. But my subclusters are not really making biological sense in that the canonical marker for ILC2s, which should be fairly rare, is lighting up everywhere, and NKs, which should be far more numerous, appear somewhat sparse.