Ease of Use: the GeneMANIA Cytoscape plugin has a user-friendly graphical user interface, which makes powerful prediction tools and data accessible to typical biologists. It can be installed using Cytoscape's menu-driven plugin manager. Upon first use, a user must download the latest version of GeneMANIA data for the organisms they are interested in. A simple graphical interface aids in this process, which can be time consuming depending on organism choice (e.g. all data for human is currently 1.4 GB compressed). However, once the database is downloaded, it does not need to be downloaded again, unless there is an update.
The plugin recognizes gene identifiers, symbols and non-ambiguous synonyms from Entrez, Ensembl, RefSeq, TAIR and Uniprot. Users can supply a mixture of symbols from different sources and the plugin will attempt to map them to the corresponding gene. Users can build their query list using the auto-completion feature, which finds genes by prefix as the user types or by pasting in large gene lists from other sources, such as text files.
The plugin can produce a prediction report that lists the details of the query list, source networks and the predicted genes. The composite interaction network used for the prediction can also be exported in standard formats, e.g. XGMML, SIF and PDF.
Customization: individual networks or entire categories can be included or excluded prior to the prediction process. Users may add their own interaction networks and gene expression profiles to this set. The plugin automatically translates the networks into an optimized matrix format, reports any unrecognized gene symbols to the user and omits the corresponding interactions.
A number of weighting methods are available to adjust the degree of influence each network has on the resulting prediction. The default weighing method (‘automatic
’) chooses between two different weighting methods depending on query list size. For longer gene lists, each network is weighted so that after the networks are combined, the query genes interact as much as possible with each other while interacting as little as possible with genes not in the list (‘query gene based
’ weighting). For shorter gene lists, an attempt is made to reproduce Gene Ontology (GO) Biological Process co-annotation patterns (Mostafavi et al.
). The two non-adaptive weighting schemes also work well on small gene lists (Mostafavi et al.
): ‘equal by network
’ weighting assigns the same weight to all networks, whereas the ‘equal by data type
’ weighting ensures each network category has the same degree of influence. Network weights can also be assigned based on how well they reproduce GO co-annotation patterns for that organism in the molecular function or cellular component hierarchies.
Provenance: each prediction is annotated with all contributing source interactions. Clicking on an interaction reveals the details about its data source and links to relevant publications, if available.
Scalability: the size of the GeneMANIA network data is limited by the amount of available memory and disk space. We recommend using a system with 4 GB of total RAM when using the default list of networks and at least 6 GB of RAM for all networks.