Skip to main content
Fig. 4 | BMC Bioinformatics

Fig. 4

From: Boolean regulatory network reconstruction using literature based knowledge with a genetic algorithm optimization method

Fig. 4

Evaluation of the optimization method: Workflow. To evaluate our optimization method, we used a gold standard network, interpreted as the true underlying biological system’s network (a). An in silico PKN (b) and a training set (c) containing a limited amount of information were generated by performing in silico experiments on the gold standard network. Using the PKN and training set as input, we started 500 independent runs of our network optimization (d), and kept the best network (with minimal value of the fitness function) obtained with each run. Among these 500 networks, the 50 best networks (according to fitness function) were kept as model networks (e). The predictive power of each model network was then evaluated by comparing its average predictions for all single node perturbations (f) to the corresponding gold standard network predictions (g). The resulting s all scores (h) measure how close model networks and gold standard network predictions are and were taken as a proxy for the quality of the optimization method

Back to article page