Presented by Jing Li, Leonard Case Jr. Professor in Engineering and Interim Chair Department of Computer and Data Sciences; and moderated by Tim Beal
Abstract: Heterogeneous networks have been widely used in modeling real-world complex systems and have been a powerful tool in studying complex biological problems. Link prediction in heterogeneous networks is one of the key computational problems. Efficient and effective algorithms for link prediction in heterogeneous networks are in great need. Furthermore, large scale network based integrative analyses that use multiple data sources have been a promising strategy for many applications in computational biology such as computational drug prediction. A key challenge in integrating multiple data sources is the lack of an extendable system that can effectively handle missing data from multiple sources. In this talk, I will go over some of our recent work in computational drug predictions based on multiple data sources. Many of the problems can be casted as missing link prediction problems on heterogeneous networks. Various approaches including random walk with restart, joint matrix-matrix decomposition, and joint tensor-matrix decomposition will be discussed.