Refine: Boolean Inference of Logical Networks

This repository contains the paper, documentation and implementation of Refine: Boolean Inference of Logical Networks, ILP vs. SAT.

Make sure Python3.9, Gurobi9.1 with an active license (https://www.gurobi.com/downloads/) and Cryptominisat5.8.0 (https://github.com/msoos/cryptominisat) are installed.

Input files are in MIDAS (experiments), SIF (network), and JSON (setup) format.

Python implementation

The implementation uses Gurobi 9.1, Cryptominisat 5.8.0 and Python 3.9.

The code can be found in:

cd src

To get an overview of the commandline arguments:

python3.9 refine.py -h

By default, the code will run the truth table version, with the flag -i for the ILP:

python3.9 refine.py ../data/caspo/ExtLiver/dataset.csv ../data/caspo/ExtLiver/pkn.sif -s ../data/caspo/ExtLiver/setup.json -i

To run the code such that it finds all models, set flag '-a':

python3.9refine.py <experiments> <network> <optional: -s setup> -i -a

To run the threshold model use flag '-t'

python3.9 refine.py <experiments> <network> <optional: -s experiments> -i -t

To run the 2-DNF model use flag '-d'

python3.9 refine.py <experiments> <network> <optional: -s experiments> -i -d