Revealing the Hidden Components in Regulatory and Signaling Networks by Integrating Proteomics, Transcriptome and Interactome Data
High throughput technologies provide a rich source for understanding signaling and regulatory mechanisms in living cells. Although a variety of "omic" data are available to probe these pathways, typically there is little overlap among the proteins and genes identified by different methods. Furthermore, these "hits" have low overlap with known pathways. Previous publications have shown that a more coherent view of the underlying biological processes can be obtained by using a network approach in which the hits from "omic" experiments are mapped onto a network of protein-protein interactions. However, given the high rates of false positives and negatives in these data, the resulting networks are frequently too large and noisy to be interpreted. One successful approach to this problem is constrained optimization to identify a subset of the "omic" hits that are connected directly or indirectly by relatively high probability interactions. This was achieved by searching for the solution to the prize-collecting Steiner tree (PCST) problem . Here, we present a web server, SteinerNet, which can effectively analyze the 'omic' data by solving the prize-collecting Steiner tree (PCST) problem to reconstruct a biologically relevant signaling pathway composed of a subset of the detected proteins/genes (terminals) through other undetected proteins present in the interactome. This method and its application on yeast pheromone response have been published in Science Signaling . Now, we aim to make this method utilized community-wide in a user-friendly way via SteinerNet. In this way, users will be able to run this approach for different organisms and to integrate distinct transcriptional and proteomic data under different conditions.
SteinerNet is free and open to non-commercial users. SteinerNet web interface is the user-friendly way to run the algorithm. In the background of the SteinerNet, three consecutive steps are running; i. processing of the input files, ii. running the branch-and-cut algorithm (DHEA, ref) to find Steiner tree, iii. post-processing of the output (see Figure below for the flowchart of the SteinerNet). At each step, a configuration file is prepared in the background. The results of the jobs will be kept for 15 days on the web server; then, they will be removed.
|Figure 1 The concept figure and the flowchart of SteinerNet. The first step is composed of input preparation. As soon as the inputs are submitted to the web interface, it converts them into configuration files and sends to the queue. Then, the prize-collecting Steiner Tree problem is solved for that specific system and the results are arranged in appropriate format to be used.|
The input to SteinerNet is a set of experimentally detected proteins/genes or specific condition-related (i.e. disease specific) proteins/genes. Each node is given a node penalty (p(v)) based on confidence in the data for that protein/gene. The algorithm then puts a premium on including nodes with the highest penalties. The protein-protein interactions are assigned costs (c(e)) equal to the negative logarithm of the probability that the interaction occurs. Minimizing the total cost of all edges in the tree and the total penalty of all nodes not contained in the tree produces a compact and biologically relevant network (see Figure 2). The resulting tree contains a subset of the original hits as well as additional nodes that represent undetected components of the response pathway. The algorithm has a single parameter, beta, which changes the relative importance of the node penalties and edge costs. All details about this method are available in ref.
|Figure 2 Finding relevant interactions as a constrained optimization problem.|
© Copyright The Fraenkel Lab 2012, Massachusetts Institute of Technology. For more information contact firstname.lastname@example.org