I have a dataset resutled from libraries by MULTI-seq and standard 10x Genomics Chromium Single Cell 3’ Reagent Kits v.3. The sample barcode is prepared using HTO approach

When doing demultiplexing, I observed that there are thousands of potential sample barcode while there are only 9 official sample barcode used.

Is this event due to contamination in barcode pool? Or is there any other reason for it?

Thanks for any answer.

  • $\begingroup$ in sc datasets you get one barcode + UMI per cell, otherwise you would have no way of telling which reads come from which cell. $\endgroup$
    – Niklas
    Commented May 10 at 13:15

1 Answer 1


You don't give much information. However, here is how 10x hashing usually goes:

First, there is the per-cell barcode that tells you which gene expression reads come from one cell. CellRanger and other software do that for you.

Then there is hashing, where you stain cells with an antibody before the actual 10x reactions. Antibody-attached oligonucleotides will later be sequenced and assigned to the cell expression matrix.

Now, CellRanger etc don't demultiplex hashtags. You have to do that manually. Meaning, you need to use software like DropletUtils to decide whether the per-cell counts per hashtag are robust to say 'yes, this cell is positive' for that antibody. There are as many hashtags as you put during the experiment. You find the hashtag counts on the tail of the expression matrix.

  • $\begingroup$ Thanks for your reply. I understand that CellRanger doesn't do demultiplex, and I need to do it manually. My problem is that I have two different types of fastq files: one for RNA fragment, and one for only sample code. The second fastq file includes sample barcode, cell barcode, and read UMI. I firstly used package deMULTIplex to obtain the matrix (cell barcode as row, sample barcode as column). After this, I observed that there are thousand of sample barcodes (means columns), while ideally there should be 9 sample barcodes for 9 samples. I'm wondering what is the reason for that? $\endgroup$
    – Tien
    Commented May 13 at 6:08
  • $\begingroup$ I also found an article in fastqc blog, it mentioned the contamination in barcode might be the reason for having strange sample barcode out of the list. So I'm wondering if it's possible to have such contamination in sample barcode pool and produce thousands of strange sample barcode? $\endgroup$
    – Tien
    Commented May 13 at 6:16
  • $\begingroup$ @Tien How many reads do you have for the 'extra' sample barcodes, and how many for the 9 barcodes you are expecting? There are always some artifact reads that end up looking like barcodes, but if they are a small minority, you can just ignore them. $\endgroup$
    – Cloudberry
    Commented May 13 at 17:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.