highly variable genes problem

[snow55]

When I did cNMF in spatial transcriptomics, in which the n_counts of one spot could be smaller than scRNA-seq cell. It usually encounter that I have to drop many spots, due to there is no highly varaible counts in them. Would you like to provide a solution that do not strictly require the counts of highly varaible counts larger than 0. That would be a great feature for the analysis of spatial transcriptomics and other technique generating sparse count matrix.

[dylkot]

for this issue I would simply lower the number of highly variable genes until you don't have this issue, or I would pre-filter spots with very low counts since many of those might not contain cells anyway.