LD highlighting

[Matthew Maher]

I recently setup+used PheWeb for the first time (nice tool!) and everything went well, except that the LD-highlighting when zoomed into a 'region' view does not seem to work.   I believe my variants (I'm using build HG38) are being properly recognized as RS#s are being assigned and displayed. 

Can anyone recommend how to diagnose what I'm doing wrong or how to correct this?  I did not see anything about LD-highlighting in the README (but perhaps I missed it).

Thanks for any assistance.

[pjvandehaar]

If you can email me the URL, I can look at it.  If not, can you send these four things:

(1) a screenshot of the LD-less LocusZoom view

(2) a screenshot of your browser dev-tools console to check for errors

(3) some screenshots of your browser dev-tools networking tab, opened to the request to "https://portaldev.sph.umich.edu/ld/...", and looking at the headers and response sections

(4) your "config.py"

[Matthew Maher]

I figured it out - sorry for the false alarm.   I had done all the data prepping on a compute cluster but then copied the results to a dedicated small instance for serving, but I had somehow failed to copy the config.py.   In quite a display of fault tolerance, despite having no config.py at all, "pheweb serve" actually works quite well (all excepting the LD highlighting)!

Again, sorry for that disturbance. 

In an unrelated matter, I made a minor mod to my copy of PheWeb to allow a config option for dealing with -log10 representation of P-values on the input - I believe I see a few people had asked for that (no doubt due to them also being regenie users) and I didn't see any such feature.   I posted a pull request for this if you're interested. 

Thanks for PheWeb (nice tool!)

-Matt

--
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/c7544871-9108-4730-9352-9edefb3ad7afn%40googlegroups.com.

[Andy Boughton]

It's worth noting that LocusZoom matches LD information based on chrom/pos/ref/alt, not rsID. The same help information for LocusZoom also applies to PheWeb:

https://my.locuszoom.org/about/#missing-ld

If you rule out the obvious things, the next would be to verify that pheweb has been correctly told to fetch build 38 data, eg for the locuszoom plot. (do requests to the portaldev LD url specify "37" or "38"?)

And, of course, to remember that by default LZ uses 1000G LD, which may be lacking many rare variants. If your rare variant is not in that dataset, there will be nothing to calculate LD relative to.

On Wednesday, April 5, 2023 at 9:15:57 AM UTC-4 mma...@broadinstitute.org wrote: