Hi,

I am trying to use singleR (or any other prediction methods) to predict cell types based on a list of known genes for each cell type. I do not have a reference dataset of cells with the signature of those genes, but I do know which genes are associated with each cell type. Is there a way to use singleR without a reference dataset but instead given a list where each item in the list is a vector with names of genes that are defining each cell type (without there expression value)?
I know that singleR is based on correlation between the data and the reference but was wondering if there is a way to tweak things such that the algorithm will take label each cell in the data based on the list of genes whose expression is highest?

Thanks!

SingleR needs a reference dataset to compute the correlations. That's fundamental to how the algorithm works, and while you can fine-tune the marker set, you will ultimately still need a reference.

If you only have sets of genes, you can use single-sample gene set methods to determine which cell-type-specific set has the strongest "activity" in each cell. AUCell is one such package that is designed for this. I find it a bit clunky to use but it seems to do a reasonable job in my few test cases.