mutcleaner.cleaners.mgnify_ddg_cleaner

Functions

clean_mgnify_ddg_dataset(pipeline)	Execute the MGnify_ddG dataset cleaning pipeline and package into a MutationDataset.
create_mgnify_ddg_cleaner([dataset_or_path, ...])	Create MGnify protein stability dataset cleaning pipeline.

Classes

MGnifyddGCleanerConfig([pipeline_name, ...])

Configuration class for MGnify protein stability dataset cleaner.

class mutcleaner.cleaners.mgnify_ddg_cleaner.MGnifyddGCleanerConfig(pipeline_name='MGnify-ddG', num_workers=16, validate_config=True, column_mapping=<factory>, type_conversions=<factory>, infer_mut_workers=16, aggregation_strategy='mean', nearest_by=<factory>, label_columns=<factory>, primary_label_column='label')

Bases: BaseCleanerConfig

Configuration class for MGnify protein stability dataset cleaner. Inherits from BaseCleanerConfig and adds MGnify-specific configuration options.

This configuration is specifically optimized for the MGnify cluster-subset schema which pre-calculates ddG values from sequence pairs without extra environmental columns (such as pH or temperature).

Attributes:

column_mappingDict[str, str]

Mapping from source columns to MutCleaner standard columns. Default targets are: ‘WT_name’ -> ‘name’, ‘wt_seq’ -> ‘wt_seq’, ‘mut_seq’ -> ‘mut_seq’, ‘ddG’ -> ‘label’.

type_conversionsDict[str, str]

Data type conversion specifications. Default forces ‘label’ to ‘float32’.

infer_mut_workersint

Number of parallel workers used for mutation inference. Set to -1 to use all CPUs.

aggregation_strategyLiteral[“mean”, “first”, “nearest”]

Strategy to aggregate labels sharing identical mutation keys. Defaults to “mean”.

nearest_byList[Tuple[str, float]]

Keep mutation by distance metrics. Unused for pure computational subsets.

label_columnsList[str]

List of score/label columns to process.

primary_label_columnstr

Primary score column for the dataset package.

Methods

from_dict(config_dict)	Create configuration object from dictionary
from_json(json_path)	Load configuration from JSON file
get_summary()	Get a human-readable summary of the configuration
merge(partial_config)	Merge partial configuration with current configuration
to_dict([exclude_callables])	Convert configuration to dictionary
to_json(json_path, **json_kwargs)	Save configuration to JSON file
validate()	Validate MGnify-specific configuration parameters.

aggregation_strategy: Literal['mean', 'first', 'nearest'] = 'mean'

column_mapping: Dict[str, str]

infer_mut_workers: int = 16

label_columns: List[str]

nearest_by: List[Tuple[str, float]]

pipeline_name: str = 'MGnify-ddG'

primary_label_column: str = 'label'

type_conversions: Dict[str, str]

validate()

Validate MGnify-specific configuration parameters.

Raises:

ValueError – If label columns are misconfigured or required mappings are missing.

Return type:

None

mutcleaner.cleaners.mgnify_ddg_cleaner.clean_mgnify_ddg_dataset(pipeline)

Execute the MGnify_ddG dataset cleaning pipeline and package into a MutationDataset.

Parameters:

pipeline (Pipeline) – The pre-constructed MGnify_ddG dataset cleaning pipeline.

Return type:

Tuple[Pipeline, MutationDataset]

Returns:

• pipeline: The executed pipeline instance. - mgnify_ddg_dataset: Standardized MutationDataset object containing data.csv, wt.fasta, and metadata.json.

mutcleaner.cleaners.mgnify_ddg_cleaner.create_mgnify_ddg_cleaner(dataset_or_path=None, config=None)

Create MGnify protein stability dataset cleaning pipeline.

This function pieces together modular basic cleaners to standardize the MGnify subsets into regularized training matrices for deep learning architectures.

Parameters:

• dataset_or_path (Union[DataFrame, str, Path, None]) – Raw dataset DataFrame or file path to the raw MGnify_ddG_Dataset.csv file.

• config (Union[MGnifyddGCleanerConfig, Dict[str, Any], str, Path, None]) – Configuration instance, parameter dictionary, or path to a JSON config.

Return type:

Pipeline

Returns:

The constructed cleaning pipeline instance populated with delayed steps.