Applicability of the SoupX ambient mRNA decontamination technique to fixed and/or multiplexed scRNAseq experiments

Question

[e.r.zakiev]

My genomics platform manager says they are using "multiplexing" of our samples. Judging by the 10x reference that means that all samples are run through the same channel. But at the end, as per the reference, they provide me with one folder per-sample outputs, like that

+-- 467
   +-- filtered_feature_bc_matrix.h5
   \-- web_summary.html
+-- 488
   +-- filtered_feature_bc_matrix.h5
   \-- web_summary.html
+-- 489
   +-- filtered_feature_bc_matrix.h5
   \-- web_summary.html

Can I still use SoupX with that kind of de-multiplexed data? Should I somehow merge them altogether? Are the same basic principles 'mRNA soup' still applicable to cells sequenced using the multiplexing technology in view of the two-tier barcoding system necessary for multiplexing?

It should be noted that I have fixed cells, that's probably relevant.

The SoupX dev points out in one of the issues on gh that it's not the right tool for that and he points to the souporcell suite, but this suite is confusing me even more tbh ("Clustering scRNAseq by genotypes" isn't what I was looking for).

[ATpoint]

I suggest you call the manager and talk to them, rather than guessing what they mean. If you get one filtered matrix per these numbers (467...) then it is not multiplexed in the sense of using something like cell hashing. Talk to them, ask for clarification and analysis help (or some heads-up), tell them which questions you want to answer, then go forward.

[e.r.zakiev]

He said it's one encapsulation per sample, so he thinks SoupX should work and the samples are actually fresh, not fixed in that case.

If they were fixed then that would have been out of the question due to multiple washes before (after?) fixing and prior to encapsulation.

What is still a bit unclear to me is the multiplexing part, which is - from what I understand - crushing all droplets from all samples together in a single mix with reads from each cell having the special multiplexing two-tier barcodes as opposed to the simple barcode in a normal sequencing. In that case all the reads from all the cells will literally have the same background 'soup', as they were all mixed together before the sequencing.

I'll come back soon with hopefully some more info