Institute of Information Science, Academia Sinica

Networks have become a key approach to understanding biological systems. Many biological networks are signed undirected networks, such as gene co-expression networks or genetic interaction networks, which consist of both positive and negative links. However, most of the previous studies either ignore the signs of links or focus on only one type of them. Considering the intrinsic differences between positive and negative links, we speculated that the interconnections among positive and negative links should show distinct features, reflecting their underlying molecular mechanisms. We have defined four different measures of link clustering coefficients to explore the topological structures of signed undirected networks. The results showed that signed molecular networks exhibited distinct structural characteristics with respect to corresponding unsigned networks. Positive links are more adhesive and tend to cluster together, while negative links are more dispersive and usually behave like bridges between positive clusters. Applying these new measures to gene co-expression networks and genetic interaction networks allow us to distinguish links with different biological contexts and identify functional modules with their inter-relationships revealed.