I have a set of fastq files. One of them is for the gene expression data and the other is feature barcode data which contain hashtag information. I was told to use cellranger count for feature barcode and then hash solo to demultiplex the hashtags. However, I saw that 10X says we can use cellranger multi to demultiplex hashtags. Are these 2 methods very different? Which one is more appropriate?
Thank you
For what it's worth, we always use CellRanger count to quantify both gene expression and hashtags / surface features. Note that count does not demultiplexing. It just tells how which barcode (that is cell) has how many counts for which surface feature. The actual demultiplexing, so deciding which cell has sufficient counts for a given feature to "trust" it or to decide which cell(s) qualify as "positive" for a feature is typically done downstream. I like hashedDrops from DropletUtils for it, but many other packages exist as well.
There are two different kinds of sample index sequences. One is found in the part of the read that is the index reads on the instrument, the other is embedded in the read itself.
Indices found in the index read are handled by whatever you use to make fastqs from bcls, so either bcl2fastq, or bcl-convert, or cellranger mkfastq on versions before 9.0 (which just has a wrapper around bcl2fastq).
If you have fastqs that are named "sample1", "sample2", then these sample indices were used, and the samples have been split.
The other kind of sample indices are in the read itself. Cellranger multi will handle these. You would be told what sample has what barcode, and the barcodes have names like BC001.
It's probably more straightforward to put samples split by index read into STARSolo. I'm not sure how STARSolo will handle reads win internal barcode. But if your fastqs are already split by sample, your samples might be already properly and totally demultiplexed, so you could use starsolo if you wanted.
