validate_lead_pvalue in study locus fillters all StydyLoci but it shouldn't

[addramir]

validate_lead_pvalue flaggs all CSs. However it is possibile to have secondary signal CSs from SuSiE with lead variants <1e-6 and it is valid. The easiest fix is to have separate thresholds for PICS (genome-wide significant) and SuSiE (for example 1e-5 or 1e-6).

[d0choa]

We must make the validate_lead_pvalue method in StudyLocus, conditional to the finemappingMethod.

[DSuveges]

Oh, I really don't like the idea: the checks performed on the aggregated data should be applicable on the full dataset. If different sources have different expectations to comply, then these checks should be performed at the source. I know doing qc is more ideal in one place, but this is not really a qc on the credible set object as a whole, rather a qc on the susie or the pics-ed credible sets.

(Not to mention how much more complicated this function would become)

[addramir]

In this case, this filter should be dropped from the QC and added to data specific postprocessing.

[d0choa]

Agreed. @addramir if you can scope the datasource-specific logic we could close this ticket

[addramir]

https://github.com/opentargets/gentropy/pull/788