High-throughput technologies generate vast amounts of experimental data, whether from transcriptomics, proteomics, or other large-scale approaches. Biologically interpreting these lists of genes or proteins represents a major challenge: beyond identifying individual components, it is necessary to account for their interactions within metabolic pathways and functional networks. Software tools capable of performing this type of analysis in an automated manner are therefore increasingly needed. It is within this context that ClueGO emerges, an application of the Cytoscape platform that selects the most representative Gene Ontology terms and biological pathways for one or several lists of genes or proteins, then organizes and visualizes them as functionally structured networks.
Recognized for its reliability, ease of use, and support for numerous species, ClueGO has gathered a large community of users. However, chaining together multiple analyses within complex workflows can prove time-consuming and error-prone when relying on manual manipulation through the graphical interface. To address this, the authors have integrated Cytoscape's cyREST programming interface into ClueGO, starting from version 2.5.0. This new feature provides programmatic access to the application's functional analyses from languages such as R, Python, or JavaScript, complementing the conventional graphical usage. The functions accessible via REST are called through an address combining the host, the port, and the desired query, with responses returned in JSON format, tab-delimited text, or binary data.
A typical ClueGO analysis proceeds in four main steps. The first consists of selecting the organism: human and mouse are included by default, and more than two hundred additional species can be downloaded. The second step corresponds to loading one or several lists of genes or proteins, with the application automatically recognizing multiple types of identifiers from the NCBI, UniProtKB, and Ensembl databases. The third step selects the representative GO terms and pathways according to predefined filters. Finally, the enrichment analysis produces a network grouping connected terms on the basis of a kappa score, which accounts for the number of genes shared between two terms and defines the functional groups. The statistical results, the network, and the various graphical representations can be retrieved via the cyREST functions.
To concretely illustrate this approach, the authors provide, as supplementary material, examples of analyses in R and Python based on gene lists from B lymphocytes and NK cells, whose results are presented as pathway networks. This work thus demonstrates how to integrate ClueGO with other Cytoscape applications and external libraries in order to perform functional analyses programmatically, within bioinformatics pipelines.