Instead of "IR_VJ_1_consensus_count", include "IR_VJ_1_consensus_umis" and "IR_VJ_1_consensus_reads"

[Ngort]

Count is a misleading term that most folks would interpret as UMIs, but when contrasting with the 10X vdj files, it actually means reads supporting a consensus. I think there should be two separate columns, one for UMIs and one for reads. At the very least, "count" should be replaced with "reads".

Also: are clonotypes and chain types determined by reads or by UMIs?

[grst]

Hi @Ngort,

we didn't come up with those names, they are part of the AIRR rearrangement standard. While the names may not be the most intuitive, they are well-defined and used across many different tools.

• consensus_count = UMI count
• duplicate_count = Read count

Also: are clonotypes and chain types determined by reads or by UMIs?

You mean, which chains are selected, if there are multiple chains per cell?
If available, chains are ranked by consensus_count, then by duplicate_count.

Cheers,
Gregor

[Ngort]

Hi @grst,
Thanks for replying. Now that you say it, it's true that they do map to consensus and duplicate counts. However, in my case they map the other way around:

`
.airr
reads | umis
878 | 2

consensus_count | duplicate_count
878 | 2
`
I'm assuming this is expected behavior.

[grst]

There we have it... I said the names were confusing.

This issue clarifies it: airr-community/airr-standards#161.
I was wrong in the comment above and your last comment and the implementation in scirpy are correct.