somatic SNV tool for ONT samples

Question

[Walter]

Hello,

I'm looking for a tool to call somatic single nucleotide variations on reads from an ONT run. I have both tumor and normal samples and have already aligned them to a reference genome.

Popular tools like mutect2 have been developed for short reads, and according to this paper it is not recommended to use tools specifically designed for short reads with long reads.

I know of tools that work with ONT reads like clair3 which is used in the epi2melabs workflow for SNPs wf-human-snp, but clair3 takes only one sample at a time, so seems for germline calling only.

Another approach would be to simply call the SNVs in both tumor and normal samples independently with clair3, and then subtract normal SNVs from tumor SNVs, but according to what I have read in a few papers and online resources like this from gatk, this is a "naive approach", and a somatic caller should use both normal and tumor samples together.

• Do you know of any tool that takes both tumor and normal long read samples to detect somatic variants (similar to mutect2 but for long reads)

• Have you used tools for short reads like mutect2 with long reads?

• Have you attempted the "naive approach" of simply subtracting SNVs?

Thanks, Walter.

[LauferVA]

The manufacturer itself recommends one of two third party tools depending on the read depth. See page 2.

Alternatively, we have found that the third-party PEPPER/DeepVariant workflow provides the best SNV calling metrics for datasets with read depths <60x, basecalled with Guppy 3.6.1 and above. We advise the third-party tool WhatsHap3 for phasing, alongside PEPPER/DeepVariant. WhatsHap is integrated into Medaka, so this additional step is unnecessary when using Medaka for SNV calling.