Divergent information about coverage and cds coverage for nextstrain/flu/h3n2/pb2

[dezordi]

Hi guys,

I'm running Nextclade v3.15.3 with the flu/h3n2/pb2 dataset (chore: release 2025-09-09--12-13-13Z).

I noticed something odd — two samples with more than 0.98 coverage show a PB2:0.0026 coverage value in the cdsCoverage column. That seems strange to me, considering that the PB2 CDS spans almost the entire segment.

The complete output file is attached.

Best,

Filipe

flu-h3n2-pb2.nextclade.tsv

[ivan-aksamentov]

Hi Filipe (@dezordi)

There are indels which caused a huge frame shift. In this case, in order to not pollute the results with many gibberish mutations, we mask the shifted region with aminoacid X, and it ends up treated as not covered.

A snippet from your tsv file:

qc.frameShifts.frameShifts
PB2:3-760
PB2:3-760

In Nextclade Web you should see a long red line of a frame shift across a large portion of CDS.

Not sure if reduction in CDS coverage should also propagate back into nuc coverage or not. Technically nucs have been read nicely, it's just translation is broken.

What do you think?

[dezordi]

Hi Ivan,

Thank you for your explanation. I agree that CDS coverage should not propagate back into nucleotide coverage. I just hadn’t paid attention to the Nextclade documentation about cdsCoverage: “Nextclade now calculates amino acid coverage of each CDS. You can find this information...” — and I was under the impression that the CDS coverage values were based on nucleotides rather than amino acids.

Thanks again for the clarification.