1. Algorithms and Computational Biology Lab, Department of Computer Science and & Information Engineering, National Taiwan University, Taiwan Modular organization of protein interaction networks Speaker: Peng-An Chen Date: 2007/03/27

2. Algorithms and Computational Biology Lab, Department of Computer Science and & Information Engineering, National Taiwan University, Taiwan Reference Feng Luo et al. Modular organization of protein interaction networks. Bioinformatics, 2007

3. Algorithms and Computational Biology Lab, Department of Computer Science and & Information Engineering, National Taiwan University, Taiwan Basic Assumption Undirected graph G=(V,E) Betweeness hypothesis –The number of shortest-paths cross an edge G-N algorithm

4. Algorithms and Computational Biology Lab, Department of Computer Science and & Information Engineering, National Taiwan University, Taiwan Definition of Module U is a sub-graph of G ind(U) –The number of edges within U outd(U) –The number of edges that connect U to G-U M U = ind(U)/outd(U) >= 1 –Or in general, M U >=S

5. Algorithms and Computational Biology Lab, Department of Computer Science and & Information Engineering, National Taiwan University, Taiwan Definition of complex module Complex module be remove by G-N Algorithm

6. Algorithms and Computational Biology Lab, Department of Computer Science and & Information Engineering, National Taiwan University, Taiwan Corollary If U, V are modules and they are adjacent then W = U union V is also a module The simple module only can be combine by –Two non-module subgraphs –One module and one non-module subgraph

7. Algorithms and Computational Biology Lab, Department of Computer Science and & Information Engineering, National Taiwan University, Taiwan Betweeness property The higher betweeness value of an edge, the more probability it is an inter-module edge The lower betweeness value of an edge, the more probability it is an intra-module edge

8. Algorithms and Computational Biology Lab, Department of Computer Science and & Information Engineering, National Taiwan University, Taiwan The agglomerative algorithm 1.Compute all-paired shortest paths 2.Reverse the G-N order 3.At the beginning, 每個 node 自己一個 subgraph 4.Select an edge and combine two subgraph if they can form a simple module 5.Repeat step 4 until no edges in the list

9. Algorithms and Computational Biology Lab, Department of Computer Science and & Information Engineering, National Taiwan University, Taiwan Result We break down a network into several smaller modules

10. Algorithms and Computational Biology Lab, Department of Computer Science and & Information Engineering, National Taiwan University, Taiwan Overlapping property What is the module contain node n?

11. Algorithms and Computational Biology Lab, Department of Computer Science and & Information Engineering, National Taiwan University, Taiwan Query modules for a protein Finding all possible U, satisfy – –M U = ind(U)/outd(U) >= 1 –L <= |U| <= H –Minimal module Applied G-N Algorithm and will not create modules which have score larger than U Score of U = f(U), f is a scoring function

12. Algorithms and Computational Biology Lab, Department of Computer Science and & Information Engineering, National Taiwan University, Taiwan Scoring function Discuss a lot last week Have no ideas use which one

13. Algorithms and Computational Biology Lab, Department of Computer Science and & Information Engineering, National Taiwan University, Taiwan Algorithm Local Search?