upper bound on # of phenotypes?

[James Warren]

We've launched a proof-of-concept deployment of PheWEB and are very impressed with the service. It is unclear, though, whether there is a practical upper bound on the maximum number of phenotypes a typical installation can support.

For example, I believe http://pheweb.sph.umich.edu/about reports it has ~1500 phenotypes. Would there be issue in having 10x that number?

Thanks!

[Peter VandeHaar]

Yeah, I think it should work.  It might need small changes:

1. Right now `pheweb matrix` reads all the files at once.  It adjusts the soft ulimit to make that possible.  Code is at https://github.com/statgen/pheweb/blob/15e916b643839de3ebae093de05fe8186acfc70f/pheweb/load/cffi/x.cpp#L345-L359 . If the hard ulimit is < 15k, you'll need to raise it, and hopefully that works.  If not, pheweb matrix could be adjusted to first make 15 matrices of 1000 phenos each, and then merge those. Or to not merge them, and query all ten partial-matrices for /variant queries.

2. The /phenotypes page, fed by generated-by-pheweb/phenotypes.json , might be ~1MB with 15k phenotypes.  If that loads too slow, it could get changed to paging.

If you're going to do all that loading, you might as well include VEP annotations.  I'm working on documenting how to do that today.

--
You received this message because you are subscribed to the Google Groups "PheWeb-UMich" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pheweb-umich...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/pheweb-umich/07431e1e-1e19-49da-8318-00a9abe6b223n%40googlegroups.com.

[Peter VandeHaar]

I added instructions on running VEP at https://github.com/statgen/pheweb#5-load-your-association-files .  A demo is at https://pheweb.org/UKB-TOPMed/pheno-filter/728.7 .

[James Warren]

Fantastic - thank you so much!