qbiocode.apps.quvine.reproducibility.graph_generator module#

Synthetic Graph Generator for QuVINE

Pre-generates and saves synthetic graphs to ensure all methods use identical instances.

Summary#

Classes:

PPIGraphGenerator

Centralized PPI preprocessing pipeline.

SyntheticGraphGenerator

Generates and saves synthetic graphs for reproducible benchmarking.

Reference#

class SyntheticGraphGenerator(output_dir, seed_manager, registry)[source]#

Bases: object

Generates and saves synthetic graphs for reproducible benchmarking.

All synthetic graphs are pre-generated once and saved to disk. Methods then load these pre-generated graphs instead of generating their own.

SYNTHETIC_FAMILIES = ['configuration_model', 'core_periphery', 'degree_corrected_sbm', 'erdos_renyi', 'grid_torus', 'heterophilic_sbm', 'modular_medium', 'modular_strong', 'powerlaw_cluster', 'random_geometric', 'random_regular', 'scale_free', 'stochastic_block_model', 'watts_strogatz_high_p', 'watts_strogatz_low_p']#
__init__(output_dir, seed_manager, registry)[source]#

Initialize graph generator.

Parameters:

  • output_dir (Path) – Root directory for saving generated graphs

  • seed_manager (SeedManager) – Seed manager for reproducible generation

  • registry (DatasetRegistry) – Dataset registry to register generated graphs

generate_all(n_nodes_list=[500, 2000, 5000], n_replicates=30)[source]#

Generate all synthetic graphs for all families, sizes, and replicates.

Parameters:

  • n_nodes_list (List[int]) – List of node counts to generate

  • n_replicates (int) – Number of replicates per family-size combination

Return type:

None

generate_single(family, n_nodes, repetition_id)[source]#

Generate a single synthetic graph instance.

Parameters:

  • family (str) – Graph family name

  • n_nodes (int) – Number of nodes

  • repetition_id (int) – Repetition index

Return type:

Tuple[Graph, Path]

Returns:

  • G (nx.Graph) – Generated graph

  • graph_path (Path) – Path where graph was saved

load_graph(graph_path)[source]#

Load a pre-generated graph from disk.

Parameters:

graph_path (Path) – Path to graph file (.graphml or .pkl)

Returns:

Loaded graph

Return type:

nx.Graph

class PPIGraphGenerator(output_dir, seed_manager, registry, processed_data_dir, registry_output_path=None)[source]#

Bases: object

Centralized PPI preprocessing pipeline.

Generates fixed disease-specific benchmark graphs per: - PPI source - disease - requested size - repetition

Outputs: - graph artifact - metadata artifact - disease node artifact - registry records

generate_single(ppi_source, disease, edge_path, requested_size, repetition_id, allow_size_expansion=False, auto_increase_size=True)[source]#
Return type:

Tuple[Graph, Path]

generate_all(ppi_sources, diseases, requested_sizes, n_repetitions, allow_size_expansion=False, auto_increase_size=True)[source]#
Return type:

None