Allele alignment issue

Community Feedback
1

Hi,

Thanks for the wonderful resource!

The issue is related to this variant https://platform.opentargets.org/variant/3_173377333_A_G

Note the direction of effect of for this credible set is negative for BMI from this specific study

https://platform.opentargets.org/credible-set/4c4290dfb79586288be204fcdc8ad9e3

This is opposite of the direction of effect for the remaining BMI sumstats studies for this variant on the variant page above, all which report the same effect allele (G).

When we look at the sumstats for this variant from the publication we see that they report the same allele for ref/alt (A):

chromosome      base_pair_location      effect_allele   other_allele    beta    standard_error  effect_allele_frequency p_value variant_id      rs_id   ci_lower        ci_upper        n 
3       173114305       A       A       -0.0190525      0.00240997      0.472431        2.67245e-15     3_173114305_A_A rs529200        -0.0237759      -0.014329       342566

Clearly an issue with the underlying data, and consideration as to whether such data points should be included in future releases of the OT Platform may be warranted.

2

Hi Vince,

Thanks for bringing this to our attention.

The credible set you mentioned is a PICS credible set derived from GWAS catalog curated top-hits data. In the curated hits data, the beta is reported as +0.019 per unit decrease in BMI, with respect to G as the affect allele. From this perspective our pipeline is correctly reporting the beta as -0.019 for BMI. It’s possible this is an error in the top-hits curation, because as you mention the direction is discordant with all other BMI studies.

We only process GWAS catalog summary statistics once they have gone through their harmonisation pipeline. The raw summary statistics for GCST90428119 are available to download, but they haven’t been harmonised. Given you mention a variant with the same ref/alt allele, it could be an underlying issue with the raw data causing harmonisation to fail.

Unfortunately these are difficult scenarios to identify at a systematic level.

We will raise this case with GWAS catalog and forward the issue here.

2 Likes