Cell annotation in Single Cell RNA seq

Question

[rogerbear]

Hi guys,

I am new to R and scRNA seq analysis. I followed Bioconductor's tutorial to analyze a scRNA seq dataset, and now I am stuck with cell annotation.

Lots of people recommend the celldex and SingleR packages for annotation, but I am working on a kidney dataset and the annotated datasets in celldex did not help me. I applied BlueprintEncodeData() and HumanPrimaryCellAtlasData() on my dataset and found cell types like "astrocytes" and "gametocytes" which do not exist in kidney.

I wonder how can I annotate my cell clusters. Any response would be appreciated.

Thanks.

[ATpoint]

SingleR (or any method) does not free you from wisely choosing your references. It will always output the most similar celltype based on a reference, so it is on you to limit to celltypes that make biological sense. Just because something is most similar does not mean it is meaningful as you experience. It is difficult to advise here because this very step you are working on is actually the one that requires most of the biological expertise, you should have some insights into the expected celltypes you are working with I guess. See whether there are papers which (at least in bulk maybe) have dissected the kidney in order to serve as references or, alternatively, perform marker gene detection between clusters and then dig into the literature and see whether you can assign a known identity to each cluster based on previous research from others. This is highly specific to both your project and the question you want to answer.

[rogerbear]

Thanks for your reply. I agree with you that I should rely on my own biological expertise to interpret these clusters based on the marker genes of each cluster.

Answer 1

[jared.andrews07]

Use a reference dataset that's more applicable. These seem promising.

[rogerbear]

thanks, i can use some of the marker as reference. And i will interpret each cluster with my own expertise.