Should cells be rows or columns when dimension reducing scRNAseq data?

Question

[jrleary]

Pretty simple question; when I generate the counts matrix for my scRNAseq data using cellranger and read the object into R, the rows are genes and the columns are cells. However I've seen recently that algorithms such as PHATE expect rows to be cells and columns to be genes. Is this actually an important distinction for dimension reduction and clustering, and if so, which algorithms expect the initial count matrix to be transposed?

[ATpoint]

I am not aware of any tool requiring this, but as always when trying new tools it is good practice to extensively read the manual and function help to be aware of any requirements the tool/function may have.