Tackling Universal Properties of Minimal Trap Spaces of Boolean Networks

Supplementary code and data for the conference article

"Tackling universal properties of minimal trap spaces of boolean networks" Sara Riva, Jean-Marie Lagniez, Gustavo Magaña López, and Loïc Paulevé Proceedings of CMSB 2023, LCNS, Springer. https://doi.org/10.48550/arXiv.2305.02442

Status

• CEGAR-based reprogramming of minimal trap spaces, using an NP and a 2QBF problem (impl-bonesis/reprogramming_cegar.py)
• CEGAR-based synthesis of locally-monotone BNs with properties of minimal trap spaces, an NP and a 2QBF problem (impl-bonesis/synthesis.py)

Requirements

• bonesis

  With pip:

  • pip install bonesis

  With conda:

  • conda install -c potassco -c colomoto bonesis

• for the synthesis: AEON.py

  With pip:

  • https://rustup.rs
  • pip install biodivine-aeon

  With conda

  • conda install -c daemontus biodivine_aeon

Usage

• to execute the CEGAR approach for marker-reprogramming problem:

python impl-bonesis/reprogramming_cegar.py instances/Moon22/L3_2013_Grieco_et_al/transition_formula.bnet '{"p":1}' 3

The first parameter is the .bnet file, the second one is the marker (in JSON), and the third is the maximum number k of components in a perturbation. To fix the marker, it is also possible to use a special node (ex: p) in the bnet file and require a marker '{"p":1}'. In the Moon22 dataset, information about uncontrollable components are available, one can use --exclude '["Apoptosis","Growth_Arrest","Proliferation"]'. If the marker is given as a component, it must be involed in the uncontrollable list.

Moreover, the markers of the Moon22 dataset are contained in the bnet files in a special component p. Concerning the Trappist dataset, the possible markers are contained in files .markers, the user can choose between them.

The dataset synthetic_bns contains random generated BNs. Also in this case, different possible markers are contained in files .markers.

• to execute the Complementary approach for marker-reprogramming problem:

python impl-bonesis/reprogramming_complementary.py  instances/Moon22/L3_2013_Grieco_et_al/transition_formula.bnet '{"p":1}' 3 

Again, uncontrollable components can be specified.

• to execute the Enumeration & Filtering approach for marker-reprogramming problem:

python impl-bonesis/reprogramming_trapspaces_naive.py  instances/Moon22/L3_2013_Grieco_et_al/transition_formula.bnet '{"p":1}' 3 

Again, uncontrollable components can be specified.

• to execute the CEGAR approach for a synthesis problem:

python impl-bonesis/synthesis.py instances/Moon22/L3_2013_Grieco_et_al/transition_formula.aeon '{"p":1}'

The first parameter is the prior knowledge graph in an .aeon file and the second one is the marker (in JSON). Another option is --maxclause which allow one to specify the maximum integer number of clauses authorised in local functions (by default is 128). The option --no-canonic leads to specify the criteria for the refinement of the under-approximation (3 for TS(y)|=M ; 1 for \exists y s.t. TS(y)!=TS(x) or 0 to obtain just a different candidate solution).