In a recent paper, Zhan, Zhang, Guan, and Zhou [Phys. Rev. E 83, 066120 (2011)] presented a modified adaptive genetic algorithm (MAGA) tailored to the discovery of maximum modularity partitions of the node set into communities in unipartite, bipartite, and directed networks. The authors claim that "detection of communities in unipartite networks or in directed networks can be transformed into the same task in bipartite networks." Actually, some tests show that it is not the case for the proposed transformations, and why. Experimental results of MAGA for modularity maximization of untransformed unipartite or bipartite networks are also discussed.