Would greatly appreciate any help I can get!
I started by integrating scRNA-seq patient tumour tissue samples following Seurat's SCT integration pipeline as described here: https://satijalab.org/seurat/articles/integration_introduction.html
I then had a Seurat object with 3 assays (SCT, Integrated, RNA).
I clustered my populations and isolated my cells of interest, subsetting to get a new Seurat object.
Re-integrated and re-clustered the isolated population of cells.
I now want to create heatmaps of gene expression in order to compare my different sub-populations i.e. compare clusters.
I saw we can use AverageExpression as described there: https://satijalab.org/seurat/articles/interaction_vignette.html#calculating-the-average-gene-expression-within-a-cluster
In the tutorial they use the RNA slot which is giving me a lot of infinite values with the warning message:
In PseudobulkExpression(object = object, pb.method = "average", :
Exponentiation yielded infinite values.
data may not be log-normed.
On GitHub some suggest to use SCT if this happens: https://github.com/satijalab/seurat/issues/3004
But when I try to do DoHeatmap with SCT I get: Error in DoHeatmap(tCellsIntegratedClusterAverages, features = c("TMEM52", : No requested features found in the scale.data slot for the integrated assay. And not just for TMEM52 but for every gene I try though they are in the 'rownames' of the SCT assay.
I have read through multiple forums on Biostars and GitHub but still not sure what the best method is - should we be using SCT, Integrated or RNA for this and if SCT then why may I be getting the error mentioned above? And if RNA, should we run NormalizeData on the RNA assay before AverageExpression or not?
Thank you in advance!!!