TR metadata updates in the gnomAD browser

[bw2]

I'm working on fixing a few misc. metadata issues on gnomAD TR pages, particularly on the PRE-MIR7-2 page:

• fixing the STRchive link
• renaming PRE-MIR7-2 to MIR7-2
• marking all 33 genotypes where one allele has 3 repeats as “low quality” on manual review (based on user-reported issue STRs: update PRE-MIR7-2 locus data broadinstitute/gnomad-browser#1788 )

I'm also thinking about whether I should update gnomAD pathogenic & normal thresholds (as well as disease associations) to match STRchive at the following loci where our thresholds currently differ:

PathogenicMin for ATXN3  differs between gnomAD (56) and STRchive (60)
PathogenicMin for C9ORF72 differs between gnomAD (31) and STRchive (251)
PathogenicMin for CNBP   differs between gnomAD (55) and STRchive (75)
PathogenicMin for DMD    is set to 59 in STRchive but not in gnomAD
PathogenicMin for FMR1   differs between gnomAD (55) and STRchive (201)
PathogenicMin for FXN    differs between gnomAD (66) and STRchive (56)
PathogenicMin for HTT    differs between gnomAD (36) and STRchive (40)
PathogenicMin for MARCHF6 differs between gnomAD (668) and STRchive (791)
PathogenicMin for NOTCH2NLC differs between gnomAD (60) and STRchive (66)
PathogenicMin for PHOX2B differs between gnomAD (24) and STRchive (26)
PathogenicMin for PRNP   differs between gnomAD (9) and STRchive (5)
PathogenicMin for SAMD12 differs between gnomAD (100) and STRchive (105)
PathogenicMin for TBP    differs between gnomAD (43) and STRchive (49)
PathogenicMin for VWA1   differs between gnomAD (3) and STRchive (1)
PathogenicMin for XYLT1  differs between gnomAD (110) and STRchive (72)
PathogenicMin for ZFHX3  differs between gnomAD (41) and STRchive (46)

These loci are currently included in gnomAD, but without listing any disease association:
PathogenicMin for AFF3   is set to 300 in STRchive but not in gnomAD
PathogenicMin for CBL    is set to 101 in STRchive but not in gnomAD
PathogenicMin for ZNF713 is set to 450 in STRchive but not in gnomAD

NormalMax for ATXN7  differs between gnomAD (33) and STRchive (27)
NormalMax for CNBP   differs between gnomAD (54) and STRchive (26)
NormalMax for EIF4A3 differs between gnomAD (11) and STRchive (12)
NormalMax for FGF14  differs between gnomAD (250) and STRchive (179)
NormalMax for NOTCH2NLC differs between gnomAD (41) and STRchive (37)
NormalMax for RILPL1 differs between gnomAD (20) and STRchive (16)
NormalMax for TBP    differs between gnomAD (42) and STRchive (40)

These loci are currently included in gnomAD, but without listing any disease association:
NormalMax for AFF3   is set to 20 in STRchive but not in gnomAD
NormalMax for CBL    is set to 79 in STRchive but not in gnomAD
NormalMax for ZNF713 is set to 22 in STRchive but not in gnomAD
NormalMax for DAB1   is set to 30 in STRchive but not in gnomAD

What do you think?

Is there anything else about gnomAD TR metadata that would be useful to fix/update right now?

[hdashnow]

Hi Ben!

Could you please share the STRchive version this data came from?

This is good timing because we've just launched locus-disease curations, basically scoring each locus based on the level of evidence in the literature.
https://strchive.org/critria/

Let me take this by category and locus:

Loci in STRchive but not in gnomAD:

PathogenicMin for DMD    is set to 59 in STRchive but not in gnomAD

DMD is tagged as a refuted locus-disease association. So that's expected that it is not in gnomAD. Or if you did add it, it should be with a big caveat. Any suggestions on making this super clear to users?
See:
DMD curation
and "Details" box on this page: https://strchive.org/loci/dmd_dmd/

PathogenicMin for AFF3   is set to 300 in STRchive but not in gnomAD
NormalMax for AFF3   is set to 20 in STRchive but not in gnomAD

PathogenicMin for CBL    is set to 101 in STRchive but not in gnomAD
NormalMax for CBL    is set to 79 in STRchive but not in gnomAD

PathogenicMin for ZNF713 is set to 450 in STRchive but not in gnomAD
NormalMax for ZNF713 is set to 22 in STRchive but not in gnomAD

We've recently curated evidence for AFF3, CBL, and ZNF713 in the literature, and all reach at least the "Moderate" level, with CBL tipping just over into "Strong". So I would recommend including them in gnomAD.

Value missing in gnomAD

NormalMax for DAB1   is set to 30 in STRchive but not in gnomAD

Looks like we have the same pathogenic min for DAB1, just no normal max in gnomAD. Let me know if you want me to dig into how we picked this value.

Not different (anymore)

PathogenicMin for C9ORF72 differs between gnomAD (31) and STRchive (251)

This was an error and was fixed in STRchive v2.16.0, see also #319 (only 3 weeks ago).

PathogenicMin for HTT    differs between gnomAD (36) and STRchive (40)

STRchive has 36 as of v2.6.0
This worries me, as that change was ~9 months ago. Is one of our files not getting updated??

That's all I have time for today, but I'm happy to go through the rest.

Warm regards,
Harriet

[bw2]

Thanks for taking a look and the corrections.
It's great to see https://strchive.org/critria/ - I've been looking forward to somebody creating a resource like this!
Do you think it would be useful to show your Classifications on gnomAD browser pages?

You're right - my comparison above was based on a pinned older commit of STRchive json. The comparison with the latest version is below.

To your other points, the DMD locus data is still visible on gnomAD, but without a disease association (removed based on your and Laurel's analysis 1+ years ago). I'm thinking of keeping that as-is given the "refuted" classification.
On the other hand, although AFF3, CBL, and ZNF713 are currently displayed without a disease association, I will update them to match STRchive.

For the DAB1 normal threshold, it's up to you.

This is the updated comparison, including some small differences I didn't share in the previous comparison. The full version is here:

PathogenicMin for AFF2   differs between gnomAD (200) and STRchive (201)
PathogenicMin for AFF3   is set to 300 in STRchive but not in gnomAD
PathogenicMin for ATXN3  differs between gnomAD (56) and STRchive (60)
PathogenicMin for CACNA1A differs between gnomAD (20) and STRchive (21)
PathogenicMin for CBL    is set to 101 in STRchive but not in gnomAD
PathogenicMin for CNBP   differs between gnomAD (55) and STRchive (75)
PathogenicMin for DMD    is set to 59 in STRchive but not in gnomAD
PathogenicMin for FMR1   differs between gnomAD (55) and STRchive (201)
PathogenicMin for FXN    differs between gnomAD (66) and STRchive (56)
PathogenicMin for MARCHF6 differs between gnomAD (668) and STRchive (650)
PathogenicMin for NOTCH2NLC differs between gnomAD (60) and STRchive (66)
PathogenicMin for PHOX2B differs between gnomAD (24) and STRchive (26)
PathogenicMin for PRNP   differs between gnomAD (9) and STRchive (5)
PathogenicMin for SAMD12 differs between gnomAD (100) and STRchive (105)
PathogenicMin for TBP    differs between gnomAD (43) and STRchive (49)
PathogenicMin for TCF4   differs between gnomAD (50) and STRchive (51)
PathogenicMin for VWA1   differs between gnomAD (3) and STRchive (1)
PathogenicMin for XYLT1  differs between gnomAD (110) and STRchive (72)
PathogenicMin for ZFHX3  differs between gnomAD (41) and STRchive (46)
PathogenicMin for ZNF713 is set to 450 in STRchive but not in gnomAD
------
NormalMax for AFF3   is set to 20 in STRchive but not in gnomAD
NormalMax for AR     differs between gnomAD (35) and STRchive (34)
NormalMax for ATXN2  differs between gnomAD (31) and STRchive (30)
NormalMax for ATXN7  differs between gnomAD (33) and STRchive (27)
NormalMax for CBL    is set to 79 in STRchive but not in gnomAD
NormalMax for CNBP   differs between gnomAD (54) and STRchive (26)
NormalMax for DAB1   is set to 30 in STRchive but not in gnomAD
NormalMax for DMD    is set to 33 in STRchive but not in gnomAD
NormalMax for EIF4A3 differs between gnomAD (11) and STRchive (12)
NormalMax for FGF14  differs between gnomAD (250) and STRchive (179)
NormalMax for NOTCH2NLC differs between gnomAD (41) and STRchive (37)
NormalMax for PABPN1 differs between gnomAD (6) and STRchive (10)
NormalMax for RILPL1 differs between gnomAD (20) and STRchive (16)
NormalMax for TBP    differs between gnomAD (42) and STRchive (40)
NormalMax for ZNF713 is set to 22 in STRchive but not in gnomAD

There's also

PathogenicMin for PABPN1 differs between gnomAD (7) and STRchive (12)
NormalMax for COMP   is set to 5 in STRchive but not in gnomAD
NormalMax for PRE-MIR7-2 is set to 4 in STRchive but not in gnomAD

but these are a special cases.

[bw2]

I've synced most thresholds. One remaining locus that I'm unsure about is NOTCH2NLC where STRchive has 66 as the minimum (https://strchive.org/loci/niid_notch2nlc/) while STRipy has 60 and 90 repeats for the 2 different disease associations (https://stripy.org/database/NOTCH2NLC).

Do you have any suggestions on this?

Also, it would be nice if STRchive could more explicitly represent situations where there's more than one disease association for a locus (eg. this locus, FMR1, etc) - both on the web page and in the data json.

[hdashnow]

Hi @bw2
NOTCH2NLC is a tricky one. There's debate in the literature as to whether there are as many as 3 diseases with the same locus and some overlapping phenotypes, while others suggest all should be grouped into a single disease. The 66 threshold is from GeneReviews. There is a mention of at least one affected individual with 60, but I'm having trouble finding more detail in the paper https://doi.org/10.1093/brain/awz372.
This is probably going to require much more digging to sort out.