FIGURE 6.

Jaccard similarity between synthetic reference and generated models with varying levels of prior knowledge. The violin plots show the distribution of Jaccard similarities achieved by MC-Boomer for synthetic models. The horizontal axis shows varying proportions of incorrect interactions randomly removed from the list of actions that MC-Boomer can choose when generating models. Removal of incorrect edges simulates the effect of prior knowledge, for example, using only interactions from a database of validated biochemical interactions. As expected, higher levels of prior knowledge lead to higher Jaccard similarities, as MC-Boomer has a higher probability of choosing correct interactions from a smaller list.