I know this question can be very tricky, but is there a rule of thumb for the selection of the threshold to filter out genes with “low” evidence?
I am using multiple ontology terms simultaneously and I would like to find a systematic way to set the thresholds and avoid doing it by myself.
Thanks a lot for your fantastic work!
Hi @frequena, and welcome to the Open Targets Community! 
There isn’t a rule of thumb that we can apply across all scores, because the scores depend on the availability of evidence. The more a target/disease is studied, the more likely there is to be strong evidence for the association, and the higher the scores.
We recommend that our users interpret the scores in a relative manner, since establishing a threshold is context-specific and very difficult! 