search

Murali-group / Beeline

BEELINE: evaluation of algorithms for gene regulatory network inference
GNU General Public License v3.0
171 stars 53 forks source link

Evaluate edges by treating them as undirected. #18

Closed adyprat closed 5 years ago

tmmurali commented 5 years ago

Is this the default or the only option? I am asking since we should ideally evaluate by treating edges as directed.

adyprat commented 5 years ago

It is not the default or the only option. I'm doing a separate evaluation treating edges as undirected. So I now have two sets of AUPRC and AUROC values, one when the network edges are treated as directed, and another when the edges are undirected. I'm doing this since some methods do not predict an edge direction. I'm planning on adding a third evaluation to check if the activation or inhibition is correctly predicted (for those methods that predict a sign for the directed edge).

tmmurali commented 5 years ago

OK, these options sound good.