Hi all,
I would like to run DEG analysis on cells within a set of clusters that are positive for my marker vs cells that are negative for my marker.
The clusters in question are 0, 1, 2, 3, 4, 6, 7, 8, 9 ,11, 12, and 13, and are all keratinocytes. I have renamed the idents of these clusters to 'keratinocytes'. I next looked for expression of my marker within the keratinocyte clusters, and then subset the cells into a 'marker-positive' group and a 'marker-negative' group. I understand that if I subset the keratinocytes, I will then be left with 2 seurat objects. How would I then perform DE between the 2 objects? Is there a better way to do this?
I would also like to know how would you strategise what cells are 'marker-positive' and what cells are 'marker-negative'? Would it be any cells that are > than the average expression can be classed as positive?
Thank you in advance!
Alex
You don't need two separate objects, I think if you change the Seurat object idents to marker_negative and marker_positive then you could use FindMarkers with idents.1 and idents.2 defined according.
If you want to do this test on each of the clusters separately, then name the idents something like marker_positive_0 and marker_negative_0 (for example).
I would also like to know how would you strategise what cells are 'marker-positive' and what cells are 'marker-negative'? Would it be any cells that are > than the average expression can be classed as positive?
I don't know what the best answer is, but given how sparse scRNA-seq data is it's not uncommon to have cells that have no reads for a particular marker (just bound to happen due to sparsity) and so I wouldn't recommend too stringent of a criteria. Ultimately how you decide on your criteria will depend on the question you are trying to answer (but which is unclear from your post).
Thank you for your advice! I will try this over the next few days. Regarding the expression, I suppose the normalisation step makes it so that any cells who's expression is above 0 can be regarded as having the marker
Log in to answer this question.
