BEELINE benchmarking of Model-X knockoffs

We made minimal modifications to BEELINE in order to benchmark various model-X knockoffs.

Our changes:

• evaluate multiple parameter settings of the same model (issue here)
• add a knockoff-based network inference method
• add datasets with protein concentration and RNA production rate revealed (small modification of BoolODE)
• benchmark FDR and undirected FDR in addition to AUPR et cetera
• Add an R script for the final plots

To run our experiments, use test_knockoffs.sh. For more info about the project, see out project main repo. For more info about BEELINE, see the original. To run our experiments:

• Install Docker and Conda
• Set up Docker images with ./initialize.sh
• Set up the BEELINE conda environment via setupAnacondaVENV.sh
• Run test_knockoffs.sh

This will place all the figures in the main directory.