Comparison Manager

GitHub Link to Code.

Comparison manager for creating data comparisons.

This module provides the ComparisonManager class that creates comparisons between different data selections for further analysis. It supports various comparison modes and automatically generates appropriate sub-comparisons.

class mdxplain.comparison.manager.comparison_manager.ComparisonManager

Manager for creating and managing data comparisons.

This class provides methods to create comparisons between different data selections (created by DataSelectorManager) for further analysis. It supports various comparison modes and automatically generates the appropriate sub-comparisons.

Supported modes:

Binary: Simple A vs B comparison
Pairwise: All possible pairs from multiple selectors
One-vs-rest: Each selector vs all others combined
Multiclass: All selectors as separate classes

Examples

Pipeline mode (automatic injection):

>>> pipeline = PipelineManager()
>>> pipeline.comparison.create_comparison(
...     "folded_vs_unfolded", "binary", "key_features",
...     ["folded_frames", "unfolded_frames"]
... )

Standalone mode:

>>> pipeline_data = PipelineData()
>>> manager = ComparisonManager()
>>> manager.create_comparison(
...     pipeline_data, "folded_vs_unfolded", "binary", "key_features",
...     ["folded_frames", "unfolded_frames"]
... )

__init__() → None

Initialize the comparison manager.

The ComparisonManager creates and manages comparison metadata. Actual data processing is done via get_comparison_data() method which uses PipelineData for memmap-safe operations.

Parameters

None: No parameters required for initialization

Returns

None: Initializes ComparisonManager instance

create_comparison(pipeline_data: PipelineData, name: str, mode: str, feature_selector: str, data_selectors: List[str] | None = None, data_selector_groups: str | List[str] | None = None) → None

Create a new comparison with specified mode and selectors.

Warning

When using PipelineManager, do NOT provide the pipeline_data parameter. The PipelineManager automatically injects this parameter.

Pipeline mode:

>>> pipeline = PipelineManager()
>>> pipeline.comparison.create_comparison("folded_vs_unfolded", "binary", "key_features", ["folded_frames", "unfolded_frames"])  # NO pipeline_data parameter

Standalone mode:

>>> pipeline_data = PipelineData()
>>> manager = ComparisonManager()
>>> manager.create_comparison(pipeline_data, "folded_vs_unfolded", "binary", "key_features", ["folded_frames", "unfolded_frames"])  # WITH pipeline_data parameter

Parameters

pipeline_dataPipelineData: Pipeline data object to store the comparison
namestr: Name for the new comparison
modestr: Comparison mode: “binary”, “pairwise”, “one_vs_rest”, “multiclass”
feature_selectorstr: Name of the feature selector to use (defines columns)
data_selectorsList[str], optional: Names of individual data selectors to compare
data_selector_groupsstr or List[str], optional: Name(s) of data selector groups to use

Returns

None: Creates ComparisonData in pipeline_data

Raises

ValueError: If comparison already exists, invalid mode, or selectors not found

Examples

>>> # Binary comparison
>>> manager.create_comparison(
...     pipeline_data, "folded_vs_unfolded", "binary", "key_features",
...     data_selectors=["folded_frames", "unfolded_frames"]
... )

>>> # Using groups
>>> manager.create_comparison(
...     pipeline_data, "cluster_comp", "one_vs_rest", "features",
...     data_selector_groups="clusters"
... )

list_comparisons(pipeline_data: PipelineData) → List[str]

List all available comparisons.

Parameters

pipeline_dataPipelineData: Pipeline data object

Returns

List[str]: List of comparison names

Examples

>>> comparisons = manager.list_comparisons(pipeline_data)
>>> print(f"Available comparisons: {comparisons}")

get_comparison_info(pipeline_data: PipelineData, name: str) → Dict[str, Any]

Get information about a comparison.

Parameters

pipeline_dataPipelineData: Pipeline data object
namestr: Name of the comparison

Returns

Dict[str, Any]: Dictionary with comparison information

Examples

>>> info = manager.get_comparison_info(pipeline_data, "conformations")
>>> print(f"Mode: {info['mode']}")
>>> print(f"Sub-comparisons: {info['sub_comparison_names']}")

remove_comparison(pipeline_data: PipelineData, name: str) → None

Remove a comparison.

Parameters

pipeline_dataPipelineData: Pipeline data object
namestr: Name of the comparison to remove

Returns

None: Removes the comparison from pipeline_data

Examples

>>> manager.remove_comparison(pipeline_data, "old_comparison")

save(pipeline_data: PipelineData, save_path: str) → None

Save all comparison data to single file.

Warning

When using PipelineManager, do NOT provide the pipeline_data parameter. The PipelineManager automatically injects this parameter.

Pipeline mode:

>>> pipeline = PipelineManager()
>>> pipeline.comparison.save('comparison.npy')  # NO pipeline_data parameter

Standalone mode:

>>> pipeline_data = PipelineData()
>>> manager = ComparisonManager()
>>> manager.save(pipeline_data, 'comparison.npy')  # pipeline_data required

Parameters

pipeline_dataPipelineData: Pipeline data container with comparison data
save_pathstr: Path where to save all comparison data in one file

Returns

None: Saves all comparison data to the specified file

Examples

>>> manager.save(pipeline_data, 'comparison.npy')

load(pipeline_data: PipelineData, load_path: str) → None

Load all comparison data from single file.

Warning

When using PipelineManager, do NOT provide the pipeline_data parameter. The PipelineManager automatically injects this parameter.

Pipeline mode:

>>> pipeline = PipelineManager()
>>> pipeline.comparison.load('comparison.npy')  # NO pipeline_data parameter

Standalone mode:

>>> pipeline_data = PipelineData()
>>> manager = ComparisonManager()
>>> manager.load(pipeline_data, 'comparison.npy')  # pipeline_data required

Parameters

pipeline_dataPipelineData: Pipeline data container to load comparison data into
load_pathstr: Path to saved comparison data file

Returns

None: Loads all comparison data from the specified file

Examples

>>> manager.load(pipeline_data, 'comparison.npy')

print_info(pipeline_data: PipelineData) → None

Print comparison information.

Warning

When using PipelineManager, do NOT provide the pipeline_data parameter. The PipelineManager automatically injects this parameter.

Pipeline mode:

>>> pipeline = PipelineManager()
>>> pipeline.comparison.print_info()  # NO pipeline_data parameter

Standalone mode:

>>> pipeline_data = PipelineData()
>>> manager = ComparisonManager()
>>> manager.print_info(pipeline_data)  # pipeline_data required

Parameters

pipeline_dataPipelineData: Pipeline data container with comparison data

Returns

None: Prints comparison information to console

Examples

>>> manager.print_info(pipeline_data)