Calling variants/SNP for single-cell RNA-seq data

Question

[QX]

Hi all,

I am trying to use GATK mutect2/haplotype (default option) for variant calling on a bam file from single-cell RNAseq data; however, the generated vcf files is just an empty file.

Does anyone know what is the issue, or any other alternatives?

Best,

Answer 1

[Nicolas Rosewick]

which scRNA-seq library type are you analyzing ?

If it's 10X it's either 5', 3' or probe-based thus the transcripts will not be covered by sequencing reads.

Also in scrna-seq read depth will be extremely low thus very difficult to make a correct SNP call

[QX]

thank you!

Answer 2

[yura.grabovska]

scAllele is another option you can try - with the heavy caveat that 10x data absolutely is not designed for variant calling. We ran it on some DIPG samples and were able to identify histone and driver mutations in some cases.

[mmorad]

Hi! I'm a PhD student and I also jump on the wagon of annotating scRNA-seq data.

I wanted to ask if you were able to verify those histone and driver mutations? Because with other tools such as SComatic and Monopogen some colleagues weren't able to obtain much results, let alone validate those mutations and since I wanted to annotate scRNA-seq I'm still wondering if it's better to keep on trying these technology-specific tools or find a way to use bulk-based methods while accepting their biases/caveats.

And what do you think about smart-seq2 data? I reckon that that format should be preferable to 10x due to how the technology works

[yura.grabovska]

We knew to expect histone mutations because DIPG is defined by H3 histone alterations. Other mutations were very patchy. In the end it wasn't something we pursued very far and doing WES on bulk was enough to get a rough allele frequency and see it change across timepoints in our specific experiment.

I haven't worked with variant calling in SS2 data so I can't really comment, but full length transcript does seem like it would be more effective.