Sample ID mapping and deleting loaded data in hail backend

[anjalijain22]

Hello,

Hope you're doing well! I have two questions regarding the recently release hail backend:

1. In our cohort, the VCF IDs of some samples do not match to the individual IDs in the seqr project and so I am wondering if its possible to provide an id mapping file (similar to how it was done in the previous infrastructure)? or would you recommend that we rename samples in our VCF from the get go?

2. Secondly, how can I remove data that is already loaded into a project, i.e., I just want to delete the variant data but not the individual/family/metadata information from the project?

Many thanks for your help!

Best,
Anjali

[hanars]

Hi Anjali,

1. It is no longer supported to have a mismatch between underlying seqr ID and VCF ID, so there is no way to upload a mapping file. One straightforward option to handle this is to rename samples in the VCF before loading , as you propose. Another is that the Individual entity in seqr has both and individual_id property, which must match the VCF ID, and an optional display_name property, which is what is actually shown in the UI if it is set. Theres no easy way to set this property, but you can do it in the django admin (/admin endpoint). So what you could do is create your individuals with the ID that will be in the VCF and then go in a change the display_name to whatever it is you want your users to see

2. This is not supported for the hail backend. Once data has been loaded, it has to stay in seqr. We plan to add support for deleting data later this year

Best,
Hana

[anjalijain22]

Hi Hana, thanks for your prompt response!
Re 1: I see. Since I have many samples that would need to be renamed, I think I'll go with renaming the samples in the VCF itself as opposed to the django interface.

Re 2: Hmm, I loaded in a dataset for testing hence was wondering how to remove the data. A follow up question: would the variant data continue to live in the hail tables even if I completely delete the project and samples?

[hanars]

Attempting to deleting projects and samples from seqr has no effect on the underlying search data, and doing is actually disabled in the seqr UI once data has been loaded because doing so will result in a variety of bugs related to search.
If you just loaded a dataset for testing and have no real data, you could just delete all the hail backend data (the entire mounted volume) and the associated test project and then start again with a clean install of the backend.

[anjalijain22]

Right, I just noticed there isn't any option to delete samples that have data loaded. Thank you for the suggestion - I think we'll do that once we're done testing.
I have another follow up question - if a sample first had an exome sequenced and later has their genome sequenced and the datasets are loaded into different seqr projects, would the sample then be counted twice potentially inflating the seqr allele counts at exomic regions? Or this wouldn't be an issue as long as we used the same individual ID?

(My apologies for bombarding you with so many questions! Really appreciate your patience!)

[hanars]

Yes, these would be counted twice, although we are planning on breaking the allele frequency out to separate WES and WGS frequencies in the near future which may help disambiguate this. Also, while it does not help with the allele counts, seqr actually now supports loading both WES and WGS data to the same project/ family so we do recommend you do that.

And no worries about all the questions! Glad you are able to use seqr and happy to help

[anjalijain22]

I see - agreed that it may be useful to have the WES and WGS counts separated out. Also, that's good to know that both WES and WGS data can be loaded into the same project now. Thanks again for answering my questions!