qbiocode.apps.quvine.reproducibility.ppi_workflow module

PPI reproducibility workflow integration for QuVINE.

Builds fixed disease-specific PPI benchmark graphs, generates fixed task splits, registers all artifacts, exports registry files, and validates readiness before methods are run.

Summary

Classes:

PPIWorkflow

End-to-end workflow for centralized PPI preprocessing and split generation.

Functions:

build_default_ppi_workflow

Convenience constructor for the centralized PPI preprocessing workflow.

Reference

class PPIWorkflow(seed_manager, registry, processed_data_dir, data_root, splits_root, registry_json_path, registry_csv_path=None)

Bases: object

End-to-end workflow for centralized PPI preprocessing and split generation.

This workflow is the orchestration layer that connects: - PPI graph preprocessing - disease seed/target preservation - connectivity repair - fixed split generation - registry export - pre-run validation

DEFAULT_TASKS = ['node_ranking', 'node_classification', 'link_prediction']

preprocess_single(ppi_source, disease, edge_path, requested_size, repetition_id, tasks=None, task_config=None, allow_size_expansion=False, auto_increase_size=True, validate=True)

Generate one fixed PPI graph instance and all requested task splits.

Return type:

Dict[str, Any]

preprocess_all(ppi_sources, diseases, requested_sizes, n_repetitions, tasks=None, task_config=None, allow_size_expansion=False, auto_increase_size=True, validate=True)

Build all configured PPI benchmark datasets and fixed task splits.

Return type:

List[Dict[str, Any]]

build_default_ppi_workflow(base_seed, processed_data_dir, data_root=PosixPath('data/ppi'), splits_root=PosixPath('splits'), registry_json_path=PosixPath('registries/ppi_dataset_registry.json'), registry_csv_path=PosixPath('registries/ppi_dataset_registry.csv'))

Convenience constructor for the centralized PPI preprocessing workflow.

Return type:

PPIWorkflow