Beta values in study colocalisation table of study locus page

[roman-hillje]

On the study locus page, study colocalisation results are shown as a table. For example, see the one below taken from study-locus/GCST002222/19_44886339_G_A

My question is about the beta values in the “Study beta” column. Are those values supposed to the beta value of the respective lead variant?

If so, then there I think something is going wrong. For example, check the highlighted row:

If I go to the study and check the beta value for the reported lead variant, I get a value of 0.51 instead of 0.48:

The beta value of 0.48 comes from variant 19_44886339_G_A, which is the variant in the left study that the lead variant colocalised with. The following beta values are found in the summary statistics:

chrom    pos ref alt      beta
19  44886339   G   A  0.479665
19  44909976   G   T  0.506941

I think that’s because the following join conditions in the “join_results_with_betas.py” script should be using the “right_” columns (chrom, pos, ref, alt) instead of the “left_”:

opentargets/genetics-colocalisation/blob/14ffc07bffed06cf40ff5220f851a1c80048d308/join_results_with_betas.py#L74

This is of course only an issue when the left lead variant and the right lead variant are not the same.

Maybe I completely misunderstood something, in which case apologies for taking your time.

[ochoa]

Hi @Roman!

I agree with you the UI is not really intuitive here and we should find a way to better represent this information.

The data is however working as expected. As you can see in the column name left_var_right_study_beta this is what the join is meant to do. When performing colocalisation there are 2 locus (leads) and 2 studies represented as right and left. In the UI you have fixed the left locus and the left study with their respective beta at the top of the page. When you look at the colocalisation table you are comparing the fixed left study-locus with multiple right study-locus.

One piece of information that colocalisation does not handle is whether the colocalising signals have the same directionality (agreeing effect sizes). You can have perfectly colocalsing signals that have opposite effects. That’s what this column aims to capture when comparing it with the beta at the top of the page.

Does this make any sense?

[roman-hillje]

Hi David,

Thank you for taking the time to explain. I didn’t expect that the shown beta value intentionally doesn’t belong to the reported lead variant, especially because the columns are right next to each other. I find it surprising that the directionality of the left locus in the right study is more informative than the (overlapping) right locus in the right study. But I’m not sure I have the expertise to assess that.

Best,
Roman