Fig. 5: Schematic representation of the workflow we developed to generate and analyse large-scale Boolean models.

Molecular interaction maps built in CellDesigner XML format are converted to executable Boolean models using the CaSQ tool. A new version of the BMA tool is then deployed on a high-performance computing cluster to identify all the models’ attractors. These attractors are filtered to keep only the steady states. Next, the filtered steady states are validated. Differentially expressed biomolecules in the models are identified using literature mining and transcriptomic data analysis. The identified biomolecule expressions are discretised and converted to a binary vector of experimentally observed Boolean values. After that, similarity scores are computed to describe the ability of the filtered steady states to reproduce the experimentally observed values. The steady states with the highest score are selected; their average vector represents the calibrated model’s state. The calibrated model can perform in silico simulations, test biological hypotheses and generate predictions.