search

huklein / demuxmix

Demultiplexing oligo-barcoded scRNA-seq data using regression mixture models
4 stars 0 forks source link

Demuxmix for combinatorial hashing #9

Open mdmanurung opened 1 year ago

mdmanurung commented 1 year ago

Hello,

In my experiment, I used two hashes per sample, which can be demultiplexed using hashedDrops by filling the 'combinations' argument. Is it possible to do the same in demuxmix? If it's not natively supported, is it reasonable to take cells labelled as multiplets but tagged by the right combinations of hashed?

Thanks in advance!

Cheers, Mikhael

huklein commented 1 year ago

Hi Mikhael,

demuxmix can demultiplex any combination of hashtags, but it currently doesn't offer a parameter similar to hashDrops' combinations. I plan to add such a parameter in the next release since you are not the first person asking for it.

In the meantime, you can use demuxmix for your experimental design with a minimal manual extra step. Only the method summary() should not be used to summarize the results, as it assumes any droplet with more than one hashtag is a multiplet. The assignment given by dmmClassify() is correct regardless of whether a single hashtag or multiple hashtags were used. When you obtain the data frame from dmmClassify(), remove all droplets classified as "uncertain" or "negative" in the 3rd column. Then assign droplets to samples based on the combination of hashtags listed in the first column.

Best, Hans