Minimum Graph Requirements

Pyntacle version 1.0

To date, Pyntacle works with unweighted and undirected graphs, which meet the following criteria:

A graph must contain at least two nodes and one edge;
Isolates (disconnected nodes) cause critical errors especially when dealing with group centrality search algorithms. For this reason, all node isolates are pruned when importing the graph from file during the Pyntacle command line execution or whe using the PyntacleImporter method in the io_stream module.
While Pyntacle will not throw errors in the presence of isolates, we still recommend the user to use the graph_initializer method in GraphUtils in the input graph of interest BEFORE running any Pyntacle task.
Note: The isolates removed by Pyntacle will be stored in the isolates graph attribute as a list of strings (see section below).
A unique identifier must be set for each node as a node name attribute. This attribute is added by default by the Pyntacle’s import methods (see the supported network file formats guide);
Two nodes must be linked by one edge only, as multigraphs are not supported;
Any instance G of the graph (representes as an igraph.Graph object) must have a name attribute that must be set with a list of strings (e.g., g[“name”] = [”Graph1”]). When importing a graph from file, the name attribute will be filled with the name of the file.
The name attribute will be used by some Pyntacle utilities, like the set operations between networks, to keep track of the original graphs which this graph is resulting from
The following special characters will throw an error if present in the network file name when importing the network via either the Pyntacle command line or via the PyntacleImporter module:
" " (white space) . - : [ ] ( ) { } =
Moreover, we recommend the user to avoid their usage in any of the igraph.Graph attribute names.

Reserved Graph Attributes

Pyntacle exchanges information with a graph through a series of reserved attributes, embedded in the igraph.Graph object. The full list of these attributes is available in the table below, along with the level at which these attributes are stored (graph, vertex or edge) and the default type that is expected by Pyntacle methods when they are recalled. They can be edited manually and their value can be changed to fit the user's need. We do recommend, however, to leave the red labeled attributes untouched, as any unforeseen modification may cause critical errors.

Attribute name	Element	Data type	Description
name	graph	list of strings	If obtained by a set operation, a graph can have multiple names, one for each originating graph.
isolates	graph	list of strings	The list of all the isolates that are removed by Pyntacle when importing the graph from any of the supported network file formats. It stores the vertex name attribute of the isolates that are removed.
sif_interaction_name	graph	string	The name of the interaction, as specified in the header of the imported SIF file, None if the graph was not imported from a SIF file.
implementation	graph	string	Allowed values: cpu, gpu, igraph. These are automatically set by Pyntacle according to topological properties and to the complexity of the graph. They drive Pyntacle to compute complex chunks of code in parallel on multi-core processors or on GPU, if available, through Numba, or on single core relying on the iGraph Python library.
name	vertex	string	Unique node name.
parent	vertex	list of strings	Useful when performing set operations between two graphs to keep track the source network of each node. Initialized to None by default.
module	vertex	string	Used when searching for communities (identified with progressive, positive integers) within a graph. It indicates the non-overlapping presence of the node in one of the found modules.
sif_interaction	edge	list of strings	The value stored in the interaction column of a SIF file (the 2^nd leftmost column) specifying the nature of the link between two nodes. Defaults to None if the graph was not imported from a SIF file. Note: Multiple interactions between nodes are collapsed into a single edge and all the interactions are stored in a list of unique values sorted lexicographically.
adjacent_nodes	edge	tuple of strings	A tuple containing the vertex name of two adjacent nodes by this edge.
weights	edge	float or integer	Weight assigned to an edge by algorithms that work with or make weighted graphs, like the PageRank index.
module	edge	string	Used when searching for communities (identified with progressive, positive integers) within a graph. It indicates the non-overlapping presence of the node in one of the found modules.