Check filtering steps in influence

[simonvh]

Unrelated to PR. But the amount of (unexpected to me) filtering is strange, this is one example. I think this is something we should take a look at and think about.

E.g. removing not-validated TFs by default will of course make the AUPRC better, since we do not have true data of those. However, if we are confident in the model I am not sure if this is the way to go

Originally posted by @Maarten-vd-Sande in #44 (comment)