SIF

SIF (Simple Interaction Format, file extension .sif) is a minimal, line-oriented plain-text format for describing the topology of a network as a list of interactions. It originated in the Cytoscape ecosystem, where it remains one of the primary formats for importing a network when it is first built, precisely because it is easy to produce by hand in a text editor or export from a spreadsheet. Its historical roots lie in molecular-biology network analysis, and its conventional vocabulary of interaction types reflects that: pp for protein-protein interactions, pd for a protein acting on DNA, and pr for a protein acting on a reaction, though the interaction-type token may be any arbitrary string.

The data model is deliberately spartan: a graph is nothing more than nodes connected by typed, directed edges, with no attributes, styling, or layout. Each line reads as a source node, a relationship (edge) type, and one or more target nodes. A line such as "nodeD typeE nodeF nodeB" is shorthand that expands into three separate edges from the same source, all sharing that type. A line with a single token declares an isolated node; self-edges (a node named as its own target) and duplicate edges between the same pair are both permitted and preserved. Delimiting follows a notable rule: if any tab characters are present, tabs separate the fields and spaces become part of node names; otherwise spaces are the delimiters and names cannot contain spaces. Node names must be unique, since identically named nodes are merged into one.

Because SIF captures only structure, per-element data is carried out of band. In classic Cytoscape workflows, node and edge properties live in separate attribute files (typically suffixed .noa and .eda), each holding one attribute as name-equals-value lines, with edges identified by the "source (type) target" convention. This separation of topology from attributes and from presentation is SIF's defining trade-off.

In the format-conversion landscape SIF sits at the simple end, alongside plain edge lists like NCOL, TSV, or TGF, and is frequently used as an interchange step into richer Cytoscape formats such as Cytoscape JSON, XGMML, or GraphML. Tools including Cytoscape, NetworkX, igraph, and pathway converters like kgml2sif read or emit it. Its strengths are legibility, trivial authoring, and easy merging of interaction datasets; its limitations are equally clear: no attributes, no styling, and no layout, so a viewer must recompute positions on every load, and only basic typed connectivity can be expressed.