Hi everyone,
I’m starting a fairly complex experiment that includes 10x single-cell, single-nucleus, and 10x multiome (snRNA-seq + scATAC-seq from the same nucleus), with control and treatment groups spread across all three modalities. I also have some patients for whom I’ve generated two modalities (sc + sn), in case that could be useful for any type of training or modeling.
I’d like to run the standard analyses (integration, differential expression, differential abundance, and differential accessibility), but I’m unsure about the best strategy to handle all modalities together.
Any recommendations or best practices would be really helpful.
Hi @dackjames unfortunately there is no guide here. This is a hard experimental setup that I would as a first step try to reconsider. It will require major effort to get it working. MultiVI can train on multi-ome data and might do well and integrate RNA and paired RNA/ATAC data. I’m not sure whether snRNA-sequencing and scRNA-sequencing will be integrated well though and sometimes this is challenging. Some ideas: it’s usually good to select shared HVGs, encoding batch key will be helpful, it’s easier if the cell-type composition is the same across experiments. Trying harmony on this might also be a good chance. Beforehand annotate the data separately to verify meaningful integration.