When querying for variants that are in Open Targets in R, I noticed that some credible set variants have a posterior probability but do not have a L2G score; is this because “all [L2G] predictions below 0.05 are filtered out” with all of the data, or does the filtering only apply to the Platform website?
All predictions below 0.05 are filtered out, from data, API and web.
The expectation is that if you didn’t find any L2G > 0.05, the model couldn’t produce even a vague prediction of the likely causal gene. There could be many reasons for that. For example, the causal gene could be beyond the cis-window under analysis or be a non-protein coding gene which are currently discarded. Future improvements in L2G will aim to improve our resolution on finding causal genes but at the moment we don’t have a reasonable prediction for such credible sets.
If strictly required, L2G < 0.05 can be recreated using gentropy and running the locus_to_gene.