Cytoscape CX2 export¶

Cytoscape reads networks in the CX2 format/). CX2 has no hyperedges, so the export asks how to project them. This notebook compares the three projections and reads the network back from CX2.

To run this notebook, create the environment from cytoscape-cx2-export.yml. See Reproduce a tutorial.

Build a graph with a hyperedge¶

In [1]:

import annnet as an

anet = an.AnnNet(directed=True)
anet.add_nodes(['A', 'B', 'C', 'D', 'E'])
anet.add_edges('A', 'B', edge_id='binary_edge', weight=1.0)
anet.add_edges('A', 'B', edge_id='parallel_edge', weight=0.65)
anet.add_edges('B', 'C', edge_id='complex_edge', directed=False, weight=0.8)
anet.add_edges(src=['B', 'C'], tgt=['D', 'E'], edge_id='reaction', directed=True, weight=2.0)
anet.attrs.update(
    'nodes',
    {
        'A': {'label': 'ligand'},
        'B': {'label': 'enzyme'},
        'C': {'label': 'cofactor'},
        'D': {'label': 'product'},
        'E': {'label': 'byproduct'},
    },
)

anet.attrs.table('edges', derived=True).select(
    ['edge_id', 'kind', 'source', 'target', 'head', 'tail']
)

import annnet as an anet = an.AnnNet(directed=True) anet.add_nodes(['A', 'B', 'C', 'D', 'E']) anet.add_edges('A', 'B', edge_id='binary_edge', weight=1.0) anet.add_edges('A', 'B', edge_id='parallel_edge', weight=0.65) anet.add_edges('B', 'C', edge_id='complex_edge', directed=False, weight=0.8) anet.add_edges(src=['B', 'C'], tgt=['D', 'E'], edge_id='reaction', directed=True, weight=2.0) anet.attrs.update( 'nodes', { 'A': {'label': 'ligand'}, 'B': {'label': 'enzyme'}, 'C': {'label': 'cofactor'}, 'D': {'label': 'product'}, 'E': {'label': 'byproduct'}, }, ) anet.attrs.table('edges', derived=True).select( ['edge_id', 'kind', 'source', 'target', 'head', 'tail'] )

Out[1]:

shape: (4, 6)

edge_id	kind	source	target	head	tail
str	str	str	str	list[str]	list[str]
"binary_edge"	"binary"	"A"	"B"	null	null
"parallel_edge"	"binary"	"A"	"B"	null	null
"complex_edge"	"binary"	"B"	"C"	null	null
"reaction"	"hyper"	null	null	["B", "C"]	["D", "E"]

Preview the network¶

The CX2 file is for Cytoscape. On a static web page, a Graphviz drawing shows the same topology.

In [2]:

from annnet.utils import plotting

plotting.plot(anet, backend='graphviz', show_edge_labels=True)

from annnet.utils import plotting plotting.plot(anet, backend='graphviz', show_edge_labels=True)

Out[2]:

Compare hyperedge export modes¶

In [3]:

for mode in ['skip', 'expand', 'reify']:
    cx2 = an.to_cx2(anet, hyperedges=mode, export_name=f'annnet-{mode}')
    edge_aspect = next(item['edges'] for item in cx2 if 'edges' in item)
    node_aspect = next(item['nodes'] for item in cx2 if 'nodes' in item)
    print(mode, 'nodes:', len(node_aspect), 'edges:', len(edge_aspect))

for mode in ['skip', 'expand', 'reify']: cx2 = an.to_cx2(anet, hyperedges=mode, export_name=f'annnet-{mode}') edge_aspect = next(item['edges'] for item in cx2 if 'edges' in item) node_aspect = next(item['nodes'] for item in cx2 if 'nodes' in item) print(mode, 'nodes:', len(node_aspect), 'edges:', len(edge_aspect))

skip nodes: 5 edges: 3
expand nodes: 5 edges: 7
reify nodes: 6 edges: 7

Round-trip through CX2¶

The reify projection adds one node for each hyperedge. AnnNet also writes a manifest into the CX2 file, so from_cx2 restores the hyperedge.

In [4]:

import json

cx2 = an.to_cx2(anet, hyperedges='reify', export_name='annnet-reified')
restored_anet = an.from_cx2(cx2)

print('restored shape:', restored_anet.shape)
print('restored hyperedges:', sorted(restored_anet.hyperedge_definitions))
print(json.dumps(cx2[:3], indent=2))

import json cx2 = an.to_cx2(anet, hyperedges='reify', export_name='annnet-reified') restored_anet = an.from_cx2(cx2) print('restored shape:', restored_anet.shape) print('restored hyperedges:', sorted(restored_anet.hyperedge_definitions)) print(json.dumps(cx2[:3], indent=2))

restored shape: (5, 4)
restored hyperedges: ['reaction']
[
  {
    "CXVersion": "2.0",
    "hasFragments": false
  },
  {
    "metaData": [
      {
        "name": "attributeDeclarations",
        "elementCount": 1
      },
      {
        "name": "networkAttributes",
        "elementCount": 1
      },
      {
        "name": "nodes",
        "elementCount": 6
      },
      {
        "name": "edges",
        "elementCount": 7
      }
    ]
  },
  {
    "attributeDeclarations": [
      {
        "nodes": {
          "label": {
            "d": "string"
          },
          "name": {
            "d": "string"
          },
          "is_hyperedge": {
            "d": "boolean"
          },
          "eid": {
            "d": "string"
          },
          "tag": {
            "d": "string"
          },
          "reaction": {
            "d": "string"
          }
        },
        "edges": {
          "interaction": {
            "d": "string"
          },
          "weight": {
            "d": "double"
          },
          "edge_id": {
            "d": "string"
          },
          "role": {
            "d": "string"
          }
        },
        "networkAttributes": {
          "__AnnNet_Manifest__": {
            "d": "string"
          },
          "name": {
            "d": "string"
          },
          "directed": {
            "d": "boolean"
          }
        }
      }
    ]
  }
]

skip, expand, and reify are explicit choices about what Cytoscape shows. skip leaves out the hyperedges, expand replaces each hyperedge with binary edges, and reify adds a node for each hyperedge.