annovarToMaf particular cases of mixed annotations

[lisemangiante]

Hi PoisonAlien,
First, thank you for this remarkable tool! I’m using it to get meaningful oncoplots from our annovar outputs. In the practice of our data, I noticed a few points in the annovarToMaf code that I would like to discuss.

1. In lines 84-85, you mentioned that in case of having “X1;X2” in the Gene.refGene annotation column, you choose to keep only X1 since this is the closest gene from the variant position I suppose. However, in the case of “exonic” variants (Func.refGene), I don’t see why this argument is relevant, since, from what I understood from annovar documentation, the variant is exonic for both gene, so the variant can be considered as equally interesting for both genes, right?

2. In lines 120+130, ann_exonic object is set for “exonic” variants (Func.refGene), right? I don’t see how “exonic;splicing” variants could be integrated in this object. Beside, in our data, all “exonic;splicing” variants (all non-synonymous) from annovar are converted into “RNA” Variant_Classification by annovar_values and not "Missense_Mutation" as we could expect from lines 130-133. Therefore, I guess “exonic;splicing” variants went within the ann_res object (line 165).

3. Finally, in line 168, ann_res object can contain “ncRNA;splicing” for instance, and in this case only the “ncRNA” annotation is kept for these variants. However, if two genes are given “X1;X2” in the Gene.refGene column, it means that X1 is ncRNA and X2 is splicing, from what I understood from annovar documentations. Splicing events is one of the items that you chose to integrate into the oncoplot by default. Here, I have the feeling that by doing that we might miss some variants that we would like to integrate within the final oncoplot.

Thank you again for your help,
Best,
Lise

[PoisonAlien]

Hello,
Thanks for the well-researched post. All your points are valid but unfortunately, there is no straightforward answer.

1. This is the issue of transcript prioritization. It is hard to guess which transcript a given mutation is affecting. Annovar seems to assign them to all possible affected transcripts and organizes them by closeness. I would suggest using VEP or better vcf2maf which will perform transcript prioritization for you.
2. Here, in exonic;splicing the first entry of exonic is chosen. exonic variants are not necessarily non-synonymous. The definition is still ambiguous to me as well. But for what it's worth, annovar exonic terms are exon_variant in SO terms that can arise from three possible sources (see the link for details). VEP classifies them as RNA, and it's the same here.
3. This is the same as your first point and concerns the transcript prioritization - whether to assign this variant to ncRNA or splice site.

I am also attaching a file here for annovar to VEP to MAF mappings, in case if you would like to check it.
I hope this helps.

annovar_to_MAF_mappings.txt