ySBGN

ySBGN is a standalone, open-source Java application that performs bidirectional translation between the GraphML dialect written by the yEd Graph Editor (yEd, from yWorks) and the standard SBGN-ML exchange format (SBGN-ML). It was developed by Irina Balaur, Ludovic Roy, Alexander Mazein, Vasundra Touré and Charles Auffray, largely at the European Institute for Systems Biology and Medicine (EISBM, Lyon), with collaborators at the Luxembourg Centre for Systems Biomedicine and NTNU, and is hosted under the community sbgn GitHub organization. The work is described in a 2022 Journal of Integrative Bioinformatics paper, "GraphML-SBGN bidirectional converter for metabolic networks."

The motivation is practical: SBGN is the de facto graphical standard for systems-biology diagrams, but the reference SBGN editors have limited layout power for very large maps. yEd is a mature, free general-purpose graph editor with strong automatic layout, and since version 3.17.1 it ships an SBGN palette. ySBGN lets biologists draw or refine large metabolic maps in yEd, then export faithful SBGN-ML for use in the standards ecosystem, or go the other way and pull existing SBGN maps into yEd for editing and relayout. Colours and annotations are preserved across the round trip.

ySBGN targets the SBGN Process Description (PD) language specifically, the sublanguage used for metabolic and reaction networks; the Activity Flow and Entity Relationship languages are not supported. Internally it uses libSBGN (libSBGN) to read and write SBGN-ML and jgrapht-core (JGraphT) to model and reconstruct the graph. A core technical problem it solves is synthesizing PD process glyphs with ports and correctly attaching consumption and production arcs, distinguishing reversible from irreversible reactions. It offers both a JavaFX GUI and a command-line JAR interface, and was designed to plug into the SBFC (SBFC) converter framework.

Limitations are documented honestly by the authors. Because conversion is done by graph reconstruction rather than XSLT, it is slow on large maps: the ReconMap 2 test case (3,763 species, 5,535 reactions) took roughly 96 seconds SBGN-to-yEd and about 17 minutes in reverse. Error and warning logging is minimal, and glyphs without a yEd SBGN-palette equivalent (for example association/dissociation and some tag orientations) are handled incompletely. It complements SBGN-editing tools such as SBGN-ED, Newt and Cytoscape rather than replacing them.