Finding Subclusters in Seurat Clusters

Question

[fouerghi20]

I have performed clustering on my Seurat object and I would like to focus on one specific cluster and find study its subclusters. To do this, I understand that you have to subset the Seurat object. However, do I also have to run FindVariableFeatures() on this subset and then RunPCA again to find the appropriate dimensions to proceed with dimensionality reduction? I ran it with both this and without it and it yielded quite different results. Not sure which process is correct. Any help is appreciated.

Answer 1

[jared.andrews07]

Yes, you will want to re-run those steps. The variable features within that cluster and between all the others are likely very different, which will have dramatic effects on the dimensionality reduction. Without doing so, I expect it'll struggle to pull out many meaningful sub-clusters.

[fouerghi20]

Thank you Andrew. I appreciate the help.

[abedkurdi10]

Hi Andrew, Do we still have to run NormalizeData and ScaleData, as well as FindVariableFeatures and RunPCA?

[jared.andrews07]

You shouldn't have to run NormalizeData again since it's just doing a log transformation and multiplying by a scaling factor, but the others, yes.

[abedkurdi10]

Thank you!