Figure 2

Outline of OrthoClust. The inputs of our pipeline are co-association networks from multiple species as well as orthology relationships. A cost function is defined based on the topology of the co-expression networks as well as orthology relationships. Each node can be in one of q possible states labeled by 1 to q. The cost function H is optimized by simulated annealing. In simulated annealing, labels are randomly assigned initially and are allowed to flip based on H. The temperature of the system is gradually lowered with a cooling factor α = 0.9. The algorithm stops if the flipping rate is low enough. The labels of nodes at the optimal configuration represent the assignment of nodes to modules. The algorithm is repeated multiple times. The resultant modules, represented by a set of Venn diagrams, could be specific or conserved.