Hi all
I have recently downloaded a publicly available scRNAseq dataset that I want to analyse. The goal will be to do some differential expression analysis between two specific cell type clusters. The raw file was in .h5ad format, which I have converted to a SingleCellExperiment object using zellkonverter. I then followed the QC, pre-processing, normalisation and feature selection steps highlighted in the OSCA handbook. I am at the point of dimensionality reduction, and have noticed that the object already contains cell type annotations.
My first question is can skip the dimensionality reduction and clustering steps as I already have defined clusters based on the cell type annotation? My second question is how can I re-create the clusters based on the existing annotation (which I trust)?
Thank you!
The h5ad might already have the PCA and UMAP/tSNE. If so you could skip the dimension reduction. If not and you want to show a dimension reduction plot like UMAP you'll need to rerun it yourself.
Also, can you expand on what you mean by recreating the clusters?
Thank you for your answer. Unfortunately, the file does not already contain the dimensionality reduction, so it sounds like I will have to do this step.
By re-creating the clusters, I mean repeat the clustering exercise so that I can get perform differential gene expression/pathway analysis between say cluster 1 and cluster 2. Hopefully, these will still be defined by the existing cell annotation. Or perhaps I am wrong - perhaps I can just go do the differential expression after dimensionality reduction as the cell annotation already exists?
For the analysis you want to do are the cell type annotations sufficient to make the comparison groups you want? If so you can skip the clustering step.
Yes, I want to compare 2 different cell types, that are already annotated. So it sounds like I should be able to skip clustering! I will give this a go. Thank you.