removeMutants code

[Emma-Manners]

Thanks very much for your recent paper, I'm reading this with interest.

I was hoping to confirm your removeMutants code, there are potentially two instances in the notebook:

if removeMutants:
%sql
delete from temp_assays where variant_id is not null or lower(description) like '%mutant%'
or lower(description) like '%mutation%' or lower(description) like '%variant%';

if removeMutants:
%sql
delete from temp_assays where variant_id is null and lower(description) like '%mutant%' or lower(description) like '%variant%'
or lower(description) like '%mutation%';

Thanks again

[greglandrum]

Hi @Emma-Manners
Yeah, I ended up with two separate forms:

1. the first is from the code which was used to identify overlapping assays in order to do statistics (this is the source of the majority of the plots and numbers in the paper). In the interests of removing as many possible sources of variation in results as possible, I just got rid of anything where there was an indication of a variant/mutation. This was potentially overly conservative, but I really wanted to be conservative for that analysis.
2. The second form is from the code used to export curated data sets. Here I was a bit less conservative and included assays where there is a variant_id. The variant_id is included in the "assay condition hash", so assays which potentially involve a variant/mutation will be included if they have a variant_id, but assays with different variant_ids will not be combined into a single data set

Does that answer the question and make sense?

[Emma-Manners]

Thanks for such a quick reply, and for the detail.

I understand 1. Exclusion of mutants.

For 2. Drop uncurated variants. I was a little unsure when I first read the paper but this code seems to select/remove only the uncurated/problematic variants i.e. where the variant_ID is null (not a curated variant) yet the description contains variant indicators? Only the assays with WT targets and curated variants are kept? If so, this makes sense and thanks for confirming.

Thanks again, ChEMBL are considering the target annotation strategy - your paper/thoughts were very useful.

[greglandrum]

For 2. Drop uncurated variants. I was a little unsure when I first read the paper but this code seems to select/remove only the uncurated/problematic variants i.e. where the variant_ID is null (not a curated variant) yet the description contains variant indicators? Only the assays with WT targets and curated variants are kept? If so, this makes sense and thanks for confirming.

Yes, that was the intent: to keep only WT targets and curated variants.

Thanks again, ChEMBL are considering the target annotation strategy - your paper/thoughts were very useful.

I'm not convinced that we've found the optimal solution here, so I'd be happy to talk about this and kick around ideas offline if it would be helpful. As you all know, I'm a giant ChEMBL fan and happy to do anything I can to help.

[Emma-Manners]

Great, thanks for the detail.

A discussion would be useful, will follow-up with you offline next week.