Thanks - that's really helpful. However, I am a bit confused as my UMAP seems to look they same whether I set return.var.only.genes = TRUE or return.var.only.genes = FALSE.
Hi,
If I set return.only.var.genes = FALSE for sctransform in Seurat, will RunPCA use all genes by default or only variable genes (as it would with return.only.var.genes = TRUE)? I am struggling to find the relevant code on their GitHub. If anyone could point me in the direction of that code, that would be really helpful.
Best wishes,
Lucy
2 answers
I think it will use all genes. This is the code from their GitHub:
if (return.only.var.genes & !conserve.memory) {
scale.data <- vst.out$y[top.features, ]
} else {
scale.data <- vst.out$y
}
This information is available in the manual. RunPCA will only use the variable features unless you specify otherwise via the features parameter. RunUMAP uses PCA reductions by default, though you can force it to run on features directly with the features parameter (probably not what you want, and will take much, much longer to run).
Log in to answer this question.