A Partitioned Approach to Protein Interaction Mapping

This is a brief description of a new program for drawing protein interactions in three-dimensional space. Our program divides nodes into three groups: biconnected subgraph in the center, terminal nodes at the outermost region, and the rest in between them. Experimental results show that the program efficiently generates a clear and aesthetically pleasing drawing of large-scale protein interaction networks and that it is an order of magnitude faster than other force-directed layouts. Protein interactions, when visualized as an undirected graph, often yield a nonplanar, disconnected graph with nodes of wide range of degrees. This paper presents a new force-directed program that divides nodes into three groups: – Group V1 is a set of terminal nodes, i.e., nodes with degree 1. – Group V2 is a set of nodes in the subgraphs separated by cutvertices, except those in the longest subgraph. – Group V3 consists of nodes which are members of neither V1 nor V2. While we find groups in the order of V1, V2, and V3, we layout them in reverse order; V3 is first positioned in the center of the sphere, V2 in the outer region of V3, and V1 in the outer region of V2 and V3. Our program was implemented in Microsoft’s C and runs on any Windows 2000/XP/Me/98/NT 4.0 system. Table 1 shows the running times on five cases, Brain (http://www.infosun.fmi. uni-passau.de/GD2001/graphC/brain.gml), Gd29 (http://www.infosun.fmi.uni-passau. de/GD2001/graphA/GD29.gml), Y2H (http://depts.washington.edu/sfields/yplm/dat a/), MIPS-G and MIPS-P (http://mips.gsf.de/proj/yeast/tables/interaction/), at each stage of partitioning nodes into three groups, finding shortest paths in three groups, and layout and drawing. The test cases of Brain and Gd29 are different from the other data not only in the size of data sets but also in the relative size of their V3. Brain has 28 nodes in V3 out of total 33 nodes and Gd29 has 129 nodes in V3 out of total 178 nodes. However, in test cases Y2H, MIPS-G, and MIPS-P the ratio of V3 to the total number of nodes is less than 1/2. Comparison of our algorithm with other forcedirected layout algorithms shows that ours is more effective for bigger graphs and for graphs that do not have excessively high portion of V3 (data not shown). If the portion of V3 is high, the improvement in speed can be marginal. Fig. 1 shows the drawings of the MIPS physical interaction data. This work was supported by the advanced backbone IT technology development program of the Ministry of Information and Communication of Korea under grant number IMT2000-C3-4. M.T. Goodrich and S.G. Kobourov (Eds.): GD 2002, LNCS 2528, pp. 370–371, 2002. c © Springer-Verlag Berlin Heidelberg 2002 A Partitioned Approach to Protein Interaction Mapping 371 Table 1. Running times at each stage on five cases. P: partitioning nodes into V1, V2 and V3, SP: finding shortest paths in all groups, LD: layout and drawing. Nodes Running times Data Edges V1 V2 V3 P SP LD total (= P + SP + LD) Brain 135 4 1 28 0.08s 0.02s 0.15s 0.25s Gd29 344 40 10 128 0.84s 0.90s 2.06s 3.80s Y2H 542 255 100 118 1.41s 0.87s 3.49s 5.77s MIPS-G 805 198 102 231 3.24s 5.16s 8.52s 16.92s MIPS-P 2372 665 289 572 56.39s 1m 18.82s 56.20s 3m 11.41s