Method in a nutshell

  • We start with a set of input genes that have a certain feature (e.g. are differentially expressed or contain certain SNPs) and set(s) of panel genes characterizing a certain condition (e.g. genes known to be related to a particular disease or miRNA target genes).
  • We then try to connect these genes using different interaction networks. The meaning of edges will differ depending on the underlying network, in BioPlex an edge indicates part of a protein complex, in mentha it represents a binary protein interaction or a protein complex, in STRING it can be a protein interaction or a functional association, and in the FunSim network it indicate that the gene products have similar functional annotation profiles.
  • In order to extract interesting subnetworks from the unfiltered/full networks, we apply different filtering criteria. One result are shortest path networks, which contain all the input and panel genes that can be linked via at most one intermediate node. More restrictive filtering is applied to yield input/panel gene networks, which only contain genes of either sets and all direct interactions between them.
  • In all filtered and unfiltered networks we then search for network clusters, i.e., parts of the network that are more densely connected than the surrounding areas.
  • All interesting subnetworks, i.e. those connecting the input and/or the panel genes, are visualized and their functional annotation is provided.
For the sake of brevity we use the term gene to refer to a gene and its gene products. That is, an interaction between two genes actually referes to the physical interaction between the two associated gene products.


The "search" tab in the main menu allows you to search across the whole website. At the moment this search is gene centered, i.e., will only find gene-related information such as parts of gene names or synonyms, database identifiers (Entrez, UniProt), or names of gene sets (e.g. Input). Search results will include any references to matching entries in the genes and networks tabs, that is, when you click on one of the network links, you will be directed to the respective network with the query gene highlighted.

Gene tables

The "genes" tab lists all the information on the different input and panel datasets. Various external databases (Entrez, UniProt, etc.) are linked via their respective identifiers and you can get all the networks were a certain gene is involved by clicking on the first cell in a table row.

Network selection

You can select the different subnetworks either in the selection area in the top of the "network" page or in the drop down menu just below. The selection area is an accordion where you can open/close different categories. For each of the networks there will be a brief summary, e.g. listing how many panel or input genes it contains, whereas the drop down area can be used for quickly browsing through various netwoks. The subnetworks are categorized into:
  • "Clusters in full network" lists all interesting networks in the unfiltered/full networks. Here, interesting mains that a cluster contains at least one of the input genes and one panel gene or two or more of the input genes.
  • "Input/panel gene networks" only contain the input and panel genes that are also present in this network, including direct interactions between them, but not allowing any intermediate nodes. Also input/panel genes present in the network but without any interactions to other input/panel genes will be listed here.
  • "Clusters in input/panel gene networks" lists all interesting network clusters that can be computed in the filtered input/panel gene network.

Network visualization

Each of the subnetworks is visualized in an interactive network browser (based on Cytoscape.js. The visualization rules for each network are the same:
  • Fill color: Input nodes are colored with a gradient representing their expression change (M_ARVDvsCTRL_tmes): from dark blue (-3.7) over white (0) to orange (+2.7). All other nodes are grey.
  • Text color: Genes with an ACM-related variant have red text, all other nodes have black text.
  • Border: Genes that are ACM-miRNA targets have a red solid border, all other nodes have no border.
  • Selection: If nodes are selected (either by clicking on them, using the search box or by selecting from a table) those elements stay normal, while all unselected elements become half-transparent.

Network exploration

  • Using your mouse you should be able to pan the network (click the mouse button on the white network background and move around) and zoom in and out.
  • When you select a node or an edge, an information popup will appear (if this doesen't appear, click again on the node/edge). Close the popup either with the top-right cross or by clicking the network background (which will remove the selection). You can select multiple nodes by holding the shift button before clicking on the node.
  • In the popup for the nodes you will get all the identifiers, GO annotation, disease associations, and potential entries in the input or panel files.
  • The GO annotations will have a three letter code in the brackets, this will tell you if it is something a human curator assigned or if a computer programm inferred the annotation, e.g. by comparing sequence similarity. The latter will have the code IEA and are of course the weakest evidence. For more information on those codes:
  • For an edge, you will get the source of the interaction, a literature reference, and mostly also a confidence score.

Network annotation

  • Below the network visualization you will find a number of tables providing annotations for the genes from the selected network.
  • The first tab will provide the gene-based information for all the genes currently selected in the network. If more than one gene is selected, you need to open/close the respective entries by clicking on them. The information here is essentially the same that you would get in the network popups, it's just a more convinient way of comparing them.
  • The second tab, which is open by default, shows some basic network information. For example, it will list the genes from the different datasets that are present in the network.
  • The next three tabs list all the GO annotations of all the genes in the network. As for all the tables, you can sort them by a column if you click on that column. The "list of genes" column will often end with '...' if there are too many entries. In those cases, just move your mouse over the text to see the complete list. If you want to expand/compact all genes lists (e.g. in order to search for a particular string), click on the table header. The button in the last column will select all the nodes that are annotated with a certain term in the network. In general the visualization and the table are linked, if you select a gene in the network the gene name will also be colored differently in the table below.
  • The last tab lists all GO BP annotations that are significantly enriched in the current network with a false discovery rate adjusted p-value < 0.05.