cellranger count output does not give all genes.

Question

[Adem80]

Dear all,

I have recently started using the cellranger from 10x for scRNA-seq data, after having used my own pipeline (with STAR alignment) for smart-seq2 data, to get the count matrix for then later analyzing with Seurat or Scanpy.

I have however faced a strange issue: not all genes appear in the features.csv file. For example, the gene Clec7a does not appear, which seems strange to me. I then checked the total number of features present in the features.csv file and it was approximately 32000, which is very different from the 53000 I used to have with my own pipeline. This seems to indicate that the genome annotation files (gtf) are somehow different.

Does 10x filter the genome annotation file in a certain way to decrease the number of genes? Is there an option to control this?

Note: I use the same reference genome version in my own pipeline as the 10x one.

Answer 1

[swbarnes2]

I assume you are talking about mouse

Here's the entry for that gene in the gtf I downloaded from ensembl

6 ensembl_havana gene 129438554 129449742 . - .gene_id "ENSMUSG00000079293"; gene_version "12"; gene_name "Clec7a"; gene_source "ensembl_havana"; gene_biotype "polymorphic_pseudogene";

See also the ensembl entry: Look at the "gene type" in the summary

http://uswest.ensembl.org/Mus_musculus/Gene/Summary?db=core;g=ENSMUSG00000079293;r=6:129438554-129449742

If you filter the gtf the way 10xGenomics suggest, that's going to be filtered away. I don't know why the gene is so labeled, when it has transcripts which are counted as protein coding.

[Adem80]

That's really helpful!

In this case, I will have to make a new reference using "cellranger mkref" without the filterings to overcome this problem.