The ResponseNet v.3 Webserver Tutorial
To go back to the ResponseNet home page click here.
This tutorial offers an overview of ResponseNet, explanation of the icons which are constant across ResponseNet pages, explanation about the Run ResponseNet form, ResponseNet output page and the sample output.
Please see our FAQ page.
This website is best viewed at a resolution of 1440×900 and above.
Contact: Esti Yeger-Lotem at estiyl@bgu.ac.il
Table of contents:
| General | Input section | Output section |
| Overview | Run ResponseNet form | Network view |
| Icons | Protein and genes to be connected | Tabs view |
| Side menu | Job details | Context menu |
| More options | Sample |
ResponseNet is a computational framework that identifies high-probability signaling and regulatory paths that connect input data sets. The input includes two weighted lists of condition-related proteins or genes, such as a set of disease-associated genes and a set of differentially expressed disease genes, and a molecular interaction network (i.e., interactome). The output is a sparse, high-probability interactome sub-network connecting the two sets that is biased toward signaling pathways. This sub-network exposes additional proteins and interactions that are potentially involved in the studied condition and their likely modes of action.
ResponseNet v.3 supports the interactomes of the yeast (Saccharomyces cerevisiae) and of human (Homo sapiens) with the addition of supporting human tissue-specific networks – New!
You may also apply ResponseNet to other organisms given that you provide a weighted interactome.
ResposeNet v.3 offers updated interactomes including human tissue-specific interactomes built using GTEx data (GTEx Consortium, 2017).
PLEASE NOTE THAT RUNNING RESPONSENET TAKES ABOUT 300 MINUTES due to the need to solve a linear programming problem with thousands of parameters.
Detailed information about the ResponseNet framework and its application are found in:
Yeger-Lotem E., Riva L., Su L.J., Gitler A., Cashikar A., King O.D., Auluck P.K., Geddie M.L., Valastyan J.S., Karger D.R., Lindquist S., Fraenkel E. 2009.
Bridging the gap between high-throughput genetic and transcriptional data reveals cellular pathways responding to alpha-synuclein toxicity. Nature Genetics 41: 316 – 323.
Icons Appearing on ResponseNet web-pages:
These icons are constant across all ResponseNet pages:
– to ResponseNet homepage from every page.
– open ResponseNet homepage in a new tab.
– to tutorial from every page.
– Run ResponseNet: Click if you want to return to the run ResponseNet form
– Login/ change User: By default, users are logged in as anonymous. If you would like your jobs to be accessed from other computers or at other times without using output links, use the login/ change user option and log in as a user, no password is required. User jobs are maintained for 30 days.
– Load previous session: Click if you are a returning user with a login that wants to load previous jobs
– Show example output: Click if you want to load a sample session.
– Tutorial: Click to get to this tutorial.
– FAQ: Click to get to the ResponseNet FAQ
– Contact us: Click to get our contact page
This page allows you to provide the following parameters:
Organism: Choose the organism you want to analyze, ResponseNet supports human, human tissues, yeast and other. Just click on the organism you want to analyze.
Interactome version: You can choose from ResponseNet offered interactomes (versioned by creation date) or choose to upload your own interactome (choose User Defined).
Proteins and genes to be connected:
This is where you provide the input source and target sets that will be connected through molecular interactions. By default, the source set is composed of proteins and the target set is composed of genes, but these can be changed.
The source and target set should be related to a common stimulus or condition.
ResponseNet proved to be particularly useful for identifying previously hidden signaling and regulatory relationships between proteins (source set) and differentially expressed genes (target set) detected upon exposing cells to a common genetic or biochemical stimulus.
Gene ID: ResponseNet currently supports yeast ORFS, Ensembl gene ids, Entrez Gene ids, and gene common names.
Source & Target Sets: Please provide a weighted list of gene ids. The format of each line is: <gene_id>\t<node type>(\t<weight>\t<capacity> optional)\n.
Node type is Protein, Gene or MIR for microRNAs.
Weight is any number between 0-1. By increasing the weight, you increase the probability that the gene will appear in the output network. If no weights are given, weights are assumed to be uniform.
Capacity is any number between 0-1. By increasing the capacity, you increase the probability that the gene will appear in the output network. If no capacity is given it is set to 1.
Source & target Lists: You can search for your protein/ gene in the textboxes, after selecting the protein/ gene it will be displayed in the list below. You may add to the source set by typing into the text box, or you can upload files with your source and target lists.
Job name: A few words that will appear as the header of the output and can be used to find the job at a later time (please do not use spaces and/or back- and forward-slash.
Job description (optional): A few words or sentences that will appear in the output, can be used to describe the input parameters.
Load interactionsfile: This button allows you to append interactions to the interactome or replace the entire interactome (if you choose User defined).
The interactome is currently composed of a list of experimentally-detected protein-protein interactions gathered from BioGRID, DIP, IntAct, and MINT and a list of experimentally-detected protein-DNA interactions gathered as described in Yeger-Lotem et al. for yeast and from TRRUST (Han et al., 2015)for human.
The format of each line is: <source node_id>\t<target node_id>\t<source node type>\t<target node type>\t<interaction type>\t<weight>(\t <direction>\t<capacity> optional)\n.
Weight is any number between 0-1. By increasing the weight, you increase the probability that the gene will appear in the output network. If no weights are given, weights are assumed to be uniform.
Capacity is any number between 0-1. By increasing the capacity you increase the probability that the gene will appear in the output network. If no capacity is given it is set to 1.
By default interactions are considered bidirectional. You may add directed edges to the interactome by ending the interaction line with 1 (e.g., ” YPL012W YLR293C Protein Protein PPI 0.763637131966 1″); the edge will be directed from the left-most protein (YPL012W) to the other (YLR293C).
Interaction List: When checked, you may add interactions to the interactome by typing into the text box. You can also separate your values using commas in the text box.
Gamma: This parameter controls the size of your output network. The default gamma value in yeast is 10 and in human is 2.5.
Typically, when gamma < 1 no solution is found, and the output network is empty; when gamma > 20 the output network is saturated and cannot include additional paths.
Motivation: ResponseNet is formulated as a minimum-cost flow optimization problem and solving it efficiently using linear programming tools. The gamma parameter controls the connectivity between the source and target sets by multiplying the amount of flux in the network. Increasing gamma will thus incorporate lower-probability paths into the output network, while decreasing gamma will remove such paths from the network.
For more details see Yeger-Lotem et al.
Capping: This parameter controls the length of your output paths. The default capping value is 0.7, the minimum value is 0 (no path) and the maximum value is 1 (long paths).
Motivation: Capping is an upper threshold on the probability of an interaction. ResponseNet is formulated as a minimum-cost flow optimization problem and the cost of an edge representing an interaction is –log(interaction probability). Therefore, interactions with probability p=1 have a cost of 0, while low-probability interactions have high costs. If no capping is used and the network contains a non-negligible number of low-cost edges, than the output network may contain long interaction paths. By using capping, we put a lower threshold on edge costs.
For more details see Yeger-Lotem et al.
Default values for gamma and capping: We recommend using the default values for each organism. In general, for an interactome of over 120,000 interactions a gamma of 10 and above is recommended, while for an interactome of about 80,000 interactions a gamma that is less than 5 is recommended. The default values for human were selected upon evaluating ResponseNet by using SPIKE maps (Paz et al, 2011). We tested capping values ranging between 0.7-1 (with steps of 0.1) and gamma values ranging between 1-50 (for 1-5 with steps of 0.5; for 5-50 with steps of 5). Recommended values are those where the output network was close to saturation, meaning that increasing gamma did not add new nodes and edges to the output, and did not increase the sensitivity of the output. These values were quite similar across all SPIKE maps. The default values for yeast were selected as described in Yeger-Lotem et al, 2009.
Randomizations: Use randomization to estimate the probability of observing a certain protein in the output by chance. You can randomize the source set, target set or both; in each case a set of equal size will be chosen randomly from the interactome.
The number of times a protein was observed will appear in the output properties tab.
Run as shortest paths: Identifies all shortest paths connecting source and target sets, without optimization.
By hitting Submit ResponseNet will identify the sub-network connecting your source and target sets.
PLEASE NOTE THAT RUNNING RESPONSENET TAKES A FEW MINUTES due to the need to solve a linear programming problem with thousands of parameters.
The output is divided into several parts The network view, the Tabs view and the Mouse tight click menu
The network view is an interactive graphical representation of the output, displayed using Cytoscape.js plugin(Franz et al., 2016). Nodes represent proteins, gene or microRNAs and edges between them represent molecular interactions.
If the target set was composed of genes then the last edge on each path represents a regulatory interaction between a transcription factor protein and its target gene.
For more details see Yeger-Lotem et al.
Network legend:
Users can retrieve various attributes of selected molecules and interactions through the tabs view on the right. The available tabs are: Properties, Layers, Gene Ontology, Summary, Downloads and Graph Options.
Properties tab:
This tab show information about selected molecules and interactions.
Molecules have the following attributes:
- HGNC symbol of the molecule.
- Ensembl Gene ID of the molecule.
- Entrez ID of the molecule.
- Type of the molecule (protein, gene, microRNA).
- OMIM Accession # if exists.
- OMIM Description if exists.
- Random Network occurrences (if randomized).
- Random Networks as PPI occurrences (if randomized).
- Random Networks as Source set occurrences (if randomized).
- Random Networks as Target set occurrences (if randomized).
- Gene Variability in the selected tissue (if calculated).
- Drug Targets of the molecule from DrugBank.
Interactions have the following information:
- Source Nodeand Target Node: the names of the connected molecules; the edge is not directed in PPIs and directed in transcription regulation interactions.
- Interaction Type: Can be PPI, i.e., protein-protein interactions, or transcription regulation, i.e., protein-DNAS interaction. Currently only PPIs are supported. The detection method is followed by a @ sign and by the name of the database reporting the interaction (e.g. Affinity Capture-MS@BIOGRID).
- Interactions Weight as defined in the interactome.
- The Flow value computed by ResponseNet, higher value means the edge is more central.
- Random Network, if randomization analysis was preformed, the number of times the edge appeared in the randomized runs.
- Detection Method(s): Provides the experimental method(s) by which the interaction was detected (e.g. Affinity Capture-Western); whether this was a low-throughput (Low) or high-throughput (High) experiment; and the database reporting the PPI (e.g. BioGrid).
Layers tab:
The user can upload another network through the Graph Options tab (ImportGraph as Layer), or to create a new layer from the existing network through the context menu after selecting nodes and edges (Create New Layer From Selection). Once a layer was added, the user may unite layers, intersect them to identify common components, or hide common components through the show symmetric differences options. To enable these operations select layers and nodes, and click the required action.
Gene Ontology tab:
This tab shows Gene Ontology annotations for selected proteins. Each GO annotation appears with its GO id, description and evidence code. GO annotations are retrieved from AMIGO using the MyGene.info web-service (Xin et al., 2015)
Summary tab:
Summarizes the parameters of the input and output sets.
Downloads tab:
The input and output network files. If randomization option was used additional files will be added to note which nodes were chosen randomly.
Graph Options tab:
Presents options that are also accessible from the context menuand more.
Context menu:
By right clicking your mouse you can perform changes to graph layout, remove nodes, and import, export and save the session.
To demonstrate the power of the ResponseNet v.3 server and the tissue-specific networks we analyzed the signaling and regulatory networks that might be involved or disrupted in muscular dystrophy. To define the source set, we gathered from the OMIM database 34 genes known to cause muscular dystrophy. To define the target set, we used 97 genes that were found to be differentially expressed in biopsies of Duchenne muscular dystrophy patients relative to healthy controls (Haslett et al., 2002). We then applied ResponseNet to find pathways that connect the two sets in the global human interactome (Fig. 1A) and in the interactome of skeletal muscle (Fig. 1B). The output subnetwork predicted by ResponseNet for skeletal muscle, which was smaller than the subnetwork predicted for global interactome (33 versus 86 connecting proteins and microRNAs) and offered several potential pathways. For example, it identified a pathway connecting the source protein DNAJB6, a chaperone that is causal gene for limb-griddle muscular dystrophy (Palmio et al., 2015), and the target gene COL1A2 that was shown to be up-regulated in muscular dystrophy patients (Haslett et al., 2002). ResponseNet positioned DNAJB6 upstream of BRMS1, which it is known to stabilize (Hurst et al., 2006), and BRMS1 upstream of HDAC2. BRMS1 and HDAC2 are part of a histone deacetylase complex (HDAC), and the predicted disruption of this pathway indeed fits with studies that linked muscular dystrophies to deregulated HDAC activity (Consalvi et al., 2011). Another pathway placed the COL3A1 gene, which was shown to be up-regulated in muscular dystrophy (Haslett et al., 2002), downstream of miR-29, whose loss was indeed connected to dystrophic muscle pathogenesis (Wang et al., 2012). ResponseNet also suggested that a problem in the SYNE2 gene which is known to cause Emery Dreifuss muscular dystrophy causes the loss of miR-29 by affecting the MYC transcription regulation complex. Notably, these pathways were not predicted in the global interactome, stressing the advantage of using context-specific networks.
Dealing with large source and target sets: Source and target sets may contain tens to thousands of proteins and genes. The ResponseNet output network obtained for such inputs is typically large and hard to view using the ResponseNet cytoscape-web interface. We recommend downloading the output network as a CyJSON file and uploading it into Cytoscape desktop for further analysis. We also recommend removing S and T from the network and using Cytoscape hierarchical layout, which often divides the network into smaller and manageable modules.
Browser compatibility:
ResponseNet v.3 has been tested on the following browsers:
| OS | Version | Chrome | Firefox | Microsoft Edge | Safari |
| Linux | Ubuntu 18.04 | 71.0 | 64 | N/A | N/A |
| Mac OSX | Mojave | 71.0 | 64 | N/A | 12 |
| Windows | 10 | 71.0 | 64 | 42 | N/A |
References:
Yeger-Lotem E, Riva L, Su LJ, Gitler AD, Cashikar AG, King OD, Auluck PK, Geddie ML, Valastyan JS, Karger DR, Lindquist S, Fraenkel E (2009) Bridging high-throughput genetic and transcriptional data reveals cellular responses to alpha-synuclein toxicity. Nature Genetics 41:316-323
Consalvi,S., Saccone,V., Giordani,L., Minetti,G., Mozzetta,C. and Puri,P.L. (2011) Histone Deacetylase Inhibitors in the Treatment of Muscular Dystrophies: Epigenetic Drugs for Genetic Diseases. Mol. Med., 10.2119/molmed.2011.00049.
Franz,M. et al.(2016) Cytoscape.js: a graph theory library for visualisation and analysis. Bioinformatics, 32, 309–311.
GTEx Consortium (2017) Genetic effects on gene expression across human tissues. Nature, 550, 204–213.
Han,H. et al.(2015) TRRUST: A reference database of human transcriptional regulatory interactions. Sci. Rep.
Haslett,J.N., Sanoudou,D., Kho,A.T., Bennett,R.R., Greenberg,S.A., Kohane,I.S., Beggs,A.H. and Kunkel,L.M. (2002) Gene expression comparison of biopsies from Duchenne muscular dystrophy (DMD) and normal skeletal muscle. Proc. Natl. Acad. Sci., 10.1073/pnas.192571199.
Hurst,D.R., Mehta,A., Moore,B.P., Phadke,P.A., Meehan,W.J., Accavitti,M.A., Shevde,L.A., Hopper,J.E., Xie,Y., Welch,D.R., et al.(2006) Breast cancer metastasis suppressor 1 (BRMS1) is stabilized by the Hsp90 chaperone. Biochem. Biophys. Res. Commun., 10.1016/j.bbrc.2006.08.005.
Palmio,J., Jonson,P.H., Evilä,A., Auranen,M., Straub,V., Bushby,K., Sarkozy,A., Kiuru-Enari,S., Sandell,S., Pihko,H., et al.(2015) Novel mutations in DNAJB6 gene cause a very severe early-onset limb-girdle muscular dystrophy 1D disease. Neuromuscul. Disord., 10.1016/j.nmd.2015.07.014.
Xin,J. et al.(2015) MyGene.info and MyVariant.info: Gene and Variant Annotation Query Services.
Wang,L., Zhou,L., Jiang,P., Lu,L., Chen,X., Lan,H., Guttridge,D.C., Sun,H. and Wang,H. (2012) Loss of miR-29 in myoblasts contributes to dystrophic muscle pathogenesis. Mol. Ther., 10.1038/mt.2012.35.