scATAC-seq doublet detection

[RegnerM2015]

Very useful and easy to use tool!

I wanted to ask if DoubletDecon can be repurposed for scATAC-seq doublet detection. Theoretically, you could run DoubletDecon on the gene activity matrix created from the scATAC-seq data.

Would this be a reasonable approach?

Thanks in advance!

[EDePasquale]

Hi,

Thanks for bringing this up, it's a great idea that would be worth exploring. I haven't tried it myself, but my biggest concern would be in the rescue step. This step is specifically designed to find unique gene expression in putative doublet clusters that would mark them as false positive predictions, but the tuning for that (min_uniq=4) would likely be problematic for scATAC-seq. To get around this and only perform the simulated doublet similarity portion of DoubletDecon, simply change PMF=FALSE.

Let me know if you play with this and tell me how it goes!

Best, Erica