Missing somatic SNP in Pacbio MAS-seq data

[YiqunCao]

Hi,

Thank you for this great tool. I'm encountering an issue while using LongcallR with --preset hifi-masseq parameter on PacBio RNA-seq data.

The tool failed to call a somatic variant that Mutect2 independently confirmed on WES data from the same sample.

Here is the corresponding entry from the DeepSomatic VCF output for that locus:
chr21 43094667 . T C 3000 RnaEdit RDS=noselect GT:GQ:DP:AF 0/1:2147483647:12986:0.14

This variant is filtered as an RnaEdit. Are there any parameters or strategies I could use to recover such variants? Thank you!

[huangnengCSU]

Hi @YiqunCao ,

LongcallR does not call somatic mutation with its current algorithm and implementation. When encountering a somatic mutation, it may observe more than two haplotypes, which is incompatible with the diploid assumption.

[YiqunCao]

Hi @huangnengCSU ,

Thank you for the clarification. In this case, the locus is diploid. Given that, it should not have been filtered out as an RNA-editing event.

Could you please advise if there is a way to:

1. Loosen the RNA-editing filter parameters?
2. Access the results before the RNA-editing filter is applied?
3. Provide a custom RNA-editing list for the tool to use?

Thank you for your help!

[huangnengCSU]

@YiqunCao

If the locus is truly diploid, there should be no other nearby heterozygous SNPs, making the somatic mutation the only candidate variant.

Otherwise, the haplotype carrying the somatic mutation would be expected to produce two fake haplotypes, one with the mutation and one without it.

In addition, the site shows an A-to-G or T-to-C substitution, which is consistent with potential A-to-G RNA editing. LongcallR requires surrounding heterozygous SNPs to infer haplotype phase and distinguish RNA editing from true SNPs. Without any such heterozygous markers, the locus can not be evaluated. So we treat the site as RNA editing.

[YiqunCao]

Hi @huangnengCSU

Thank you for the detailed explanation. I understand the logic.

Yes, there are no other nearby heterozygous SNPs. Given that this somatic mutation is clinically significant, I am concerned that LongcallR's RNA-editing filter might be too stringent for this specific case.

Is there a way to adjust the sensitivity of the RNA-editing filter, or to disable it for certain regions? Being able to provide a list of known, clinically important variants to bypass this filter would be extremely helpful.

Thank you for considering this.

[huangnengCSU]

@YiqunCao

If I provide a option that disables entire RNA-editing filtering, does it work for you?

[YiqunCao]

Hi @huangnengCSU

Yes, that would also be very helpful. Thank you!

[huangnengCSU]

@YiqunCao

There is a branch called disable_rna_edit_filter which provides --rna-edit-filter false to disable the filtering.