Calling somatic mutations

[hukai916]

Hi there,

According to the following FAQ, DeepVariant is not supposed for somatic variant calling:

https://github.com/google/deepvariant/blob/r1.8/docs/FAQ.md#can-i-use-deepvariant-for-somatic-non-germline-calling

Instead, DeepSomatic should be adopted. However, our study involves RNA-seq data and it seems that DeepSomatic only supports DNA-seq data. Therefore, I am wondering if DeepVariant can be repurposed to call somatic mutations using RNA-seq data? Thanks,

--Kai

[danielecook]

@hukai916 you could try to use DeepSomatic with the WES model. Be sure to set the --split_skip_reads=True flag when you run make_examples.

We would be interested in experimenting with RNA-seq for somatic calling, but unfortunately, I am not aware of any high quality training data that we could make use of.

We have generally relied on data from the seqc2 consortium. If you look at the figure on that website it suggests a Study Name of RNA, but I have not seen any RNA-seq data available on the FTP website nor on SRA.

[hukai916]

Thanks for the feedback, we will test out DeepSomatic and compare the results with Mutect2, and report back.

[danielecook]

Thank you - we are open to feedback.