Search

Link prediction aims to predict whether two nodes in a network are likely to get connected. Motivated by its applications, e.g., in friend or product recommendation, link prediction has been extensively studied over the years. Most link prediction methods are designed based on specific assumptions that may or may not hold in different networks, leading to link prediction methods that are not generalizable. Here, we address this problem by proposing general link prediction methods that can capture network-specific patterns. Most link prediction methods rely on computing similarities between between nodes. By learning a γ-decaying model, the proposed methods can measure the pairwise similarities between nodes more accurately, even when only using common neighbor information, which is often used by current techniques.

Protein-protein interaction (PPI) networks are valuable biological data source which contain rich information useful for protein function prediction. The PPI network data obtained from high-throughput experiments is known to be noisy and incomplete. In the literature, common neighbor, clustering, and classification-based approaches have been proposed to improve the performance of protein function prediction by modeling PPI data as a graph. These approaches exploit the fact that protein shares function with other proteins directly interacting with it. In this paper we have experimented an alternative approach by exploiting the notion that two proteins share a function if they have a well defined group of directly or indirectly connected common neighbors. The experiments conducted on variety of PPI network datasets show that the proposed approach improves protein function prediction accuracy over existing approaches.

Protein-protein interaction (PPI) networks are valuable biological data source which contain rich information useful for protein function prediction. The PPI network data obtained from high-throughput experiments is known to be noisy and incomplete. In the literature, common neighbor, clustering, and classification-based approaches have been proposed to improve the performance of protein function prediction by modeling PPI data as a graph. These approaches exploit the fact that protein shares function with other proteins directly interacting with it. In this paper we have experimented an alternative approach by exploiting the notion that two proteins share a function if they have a well defined group of directly or indirectly connected common neighbors. The experiments conducted on variety of PPI network datasets show that the proposed approach improves protein function prediction accuracy over existing approaches.

Disease-gene networks are bipartite networks that connect diseases to the genes with which they are associated. Due to the absence of tools to analyze bipartite networks, however, they are usually analyzed by "projecting" the bipartite network into a unipartite one by taking one side of the bipartition as the nodes and connecting two nodes if they share a common neighbor in the bipartite network. This projection, however, loses information, such as exactly which groups of diseases share a common gene.Here, we present an alternative method to analyze these networks by using hypergraph techniques. We review several metrics from previous work and demonstrate their ability to find related diseases in disease-gene network.