0
$\begingroup$

I have a question related to processing of Single cell RNA-seq data originating from non 10x platform. I have worked with data that originated from 10x platform and I parse it through cell ranger pipeline and then put it through Seurat.

As an example the dataset in this GSE156644 repository are not being processed through Cell ranger pipeline and I get an error that indicates "single-end read found" another example is of this GSE115235, in the later I have bunch of files that look like bulk RNA-seq.

If you check the GSE115235 it has bunch of files that are not cellranger compatible. However, they are from single cell experiment. I find it strange that NCBI accepts such submission and the reviewers don't even pay attention to what has been submitted.

I guess, that these are SE RNA seq data that can be aligned using STAR or Bowtie and then merged in some way to get per cell per sample expression.

Along similar lines, how do I deal with 2 fastq files for a scRNA seq run? example https://www.ebi.ac.uk/ena/browser/view/PRJNA602526

This link has scRNA seq files that look paired end (R1 & R2 but NO I1). I am bit confused as cell ranger count option requires the folder to have 3 files (I1,R1,R2). How can I run cell ranger with R1 & R2? Or do I need to use another aligner?

I will send an email to sra and also to the authors, but any other help to navigate through these datasets and process them is deeply appreciated.

Thank you.

$\endgroup$
1
  • 1
    $\begingroup$ 1. Cell ranger does not require I1 and it was a bad idea for them to ever support that. You can directly run cellranger count on demultuplexed fastq files. 2. Salmon alevin or STAR-solo would be examples of common alternatives. Salmon alevin is in fact much faster and likely better. $\endgroup$
    – Devon Ryan
    Jan 24 at 6:19

1 Answer 1

2
$\begingroup$

Modified from @devon-ryan's comments:

The couple of samples I looked at have the sequence from read 1 embedded in their header. I1 isn't needed, they're already demultiplexed. You could feed those into STAR or similar tools, but processing them will be more of a pain since most tools expect the original unprocessed version of what they uploaded.

Modified from atpoint's comments:

Single-cell data is not really useful as "bulk" as the notable amount of PCR during the library prep creates quite a bias that without the UMI information cannot be corrected. I'd email the authors and ask for the R1 files. Single-cell uploads at NCBI tend to be a mess, that's unfortunately how it is. Without R1 data are almost useless.

Those file format issues are strange, but it is what it is. Reviewers do not pay attention to method sections, choice of tools and data submission most of the time in my experience, especially not when the paper is biology-focused rather than a computational-focused one. Even more, you often do not find the suitable metadata for a proper analysis such as information about batches etc. The only choice you have is emailing the authors as noted above.

If by "two files" you mean that the same library (=sample) was sequenced over multiple lanes (very common to do that since required depth is often larger than what a single lane can provide) then merge the fastq files, so all R1s into one and all R2 into one file.

$\endgroup$
0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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