Inspecting the code, the error happens because of this line

The query build process is quite complicated, also because apparently we have this criterion that whenever the user specifies a clinsig (clinsign_filter) plus secondary criteria (secondary_filter), then the query becomes:

mongo_query["$or"] = [ {"$and": secondary_filter}, clinsign_filter]

Which will return any variant matched by the first $or or the second.

In the case of the bug the query becomes the following:

....{'$or': 
[
  {'$and': [{'clnsig': {'$exists': True}}, {'clnsig': {'$ne': None}}]}, {'clnsig': {'$elemMatch': {'$and': [{'$or': [{'value': {'$in': [3, 
  'Likely benign']}}, {'value': re.compile('Likely benign')}]}, 
   
{'revstat':re.compile('mult|multiple_submitters|single|single_submitter|exp|reviewed_by_expert_panel|guideline|practice_guideline')}]}}}

]}

All variants, regardless of the confidence level are returned because of the first element of the query .. 🤔

Bug is perhaps a strong word since the interface doesn't clearly specify the priority for the "clinvar hits" checkbox. But it is sure hard to figure out a simple expression describing the design.

All three have use cases: being able to fully select/deselect "confident" annotations seems like a good feature, as does selecting any tag. Perhaps that could be an option for the multi select instead/later? Either a button that just selects all, or an "All" option like for the chromosomes?

All three have use cases: being able to fully select/deselect "confident" annotations seems like a good feature, as does selecting any tag. Perhaps that could be an option for the multi select instead/later? Either a button that just selects all, or an "All" option like for the chromosomes?

I would call it a bug, because if you filter:

• variants with clinsig
• likely benign
• high confidence

You would expect no variants returned, not all the variants with any clinsig?

But I also understand the design, which is: return variants with clinsig no matter what. But I start to dislike this criterion more and more because you don't really get the chance to decide for yourself, like: I want to really make sure no variants are present if I filter by those 3 criteria.

Regarding including "All" in the clinsig multiselect, isn't it the same as not selecting anything at all? I think I would rather keep it as it is (with the ClinVar hits checkbox), to be able to easily exclude those variants that have no ClinVar annotations

Right, two pictures here, one as it is, one as one might want to refactor it.

As it is, I would expect the above to return at least any high significance hits, but I feel a conflict in design between "clinvar hits"and "likely benign". It's kind of obvious the one named "clinsig confidence always returned" is inclusive. The others less clear.

I do agree that designs like "always" are really hard to get into a flexible system. It would make sense to refactor to retain good functionality for the clinical filter, but make the interface option a bit less strained. Meeting now! 😓