abaco.batch_correction module

class abaco.batch_correction.ConQur(batch_cols, covariate_cols, reference_batch, quantiles=(0.05, 0.5, 0.95), logistic_kwargs=None, quantile_kwargs=None)

Bases: TransformerMixin, BaseEstimator

Conditional Quantile Regression (ConQuR) batch correction transformer.

Parameters:

• batch_cols (list of str) – List of batch column names.

• covariate_cols (list of str) – List of covariate column names.

• reference_batch (dict) – Dictionary specifying reference batch values for each batch column.

• quantiles (tuple of float, optional) – Quantiles to use for quantile regression, by default (0.05, 0.5, 0.95).

• logistic_kwargs (dict, optional) – Keyword arguments for LogisticRegression.

• quantile_kwargs (dict, optional) – Keyword arguments for QuantileRegressor.

_logit_models

Fitted logistic regression models for zero-mass.

Type:

dict

_quantile_models

Fitted quantile regression models for nonzero values.

Type:

dict

_col_order

Order of columns used in the model.

Type:

list

_feature_cols

List of feature columns.

Type:

list

fit(df, y=None)

set_fit_request(*, df: bool | None | str = '$UNCHANGED$') → ConQur

Configure whether metadata should be requested to be passed to the fit method.

Note that this method is only relevant when this estimator is used as a sub-estimator within a meta-estimator and metadata routing is enabled with enable_metadata_routing=True (see sklearn.set_config()). Please check the User Guide on how the routing mechanism works.

The options for each parameter are:

• True: metadata is requested, and passed to fit if provided. The request is ignored if metadata is not provided.

• False: metadata is not requested and the meta-estimator will not pass it to fit.

• None: metadata is not requested, and the meta-estimator will raise an error if the user provides it.

• str: metadata should be passed to the meta-estimator with this given alias instead of the original name.

The default (sklearn.utils.metadata_routing.UNCHANGED) retains the existing request. This allows you to change the request for some parameters and not others.

Added in version 1.3.

Parameters:

df (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for df parameter in fit.

Returns:

self – The updated object.

Return type:

object

set_transform_request(*, df: bool | None | str = '$UNCHANGED$') → ConQur

Configure whether metadata should be requested to be passed to the transform method.

Note that this method is only relevant when this estimator is used as a sub-estimator within a meta-estimator and metadata routing is enabled with enable_metadata_routing=True (see sklearn.set_config()). Please check the User Guide on how the routing mechanism works.

The options for each parameter are:

• True: metadata is requested, and passed to transform if provided. The request is ignored if metadata is not provided.

• False: metadata is not requested and the meta-estimator will not pass it to transform.

• None: metadata is not requested, and the meta-estimator will raise an error if the user provides it.

• str: metadata should be passed to the meta-estimator with this given alias instead of the original name.

The default (sklearn.utils.metadata_routing.UNCHANGED) retains the existing request. This allows you to change the request for some parameters and not others.

Added in version 1.3.

Parameters:

df (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for df parameter in transform.

Returns:

self – The updated object.

Return type:

object

transform(df)

class abaco.batch_correction.PLSDA(ncomp=1, keepX=None, tol=1e-06, max_iter=500)

Bases: object

Partial Least Squares Discriminant Analysis (PLSDA) implementation.

Parameters:

• ncomp (int, optional) – Number of components to extract, by default 1.

• keepX (list of int, optional) – Number of variables to keep for each component (for sparsity).

• tol (float, optional) – Convergence tolerance, by default 1e-6.

• max_iter (int, optional) – Maximum number of iterations, by default 500.

t_

X scores.

Type:

numpy.ndarray

u_

Y scores.

Type:

numpy.ndarray

a_

X loadings.

Type:

numpy.ndarray

b_

Y loadings.

Type:

numpy.ndarray

iters_

Number of iterations per component.

Type:

list

exp_var_

Explained variance per component.

Type:

list

fit(X, Y)

abaco.batch_correction.correctBMC(data, sample_label, batch_label, exp_label)

This function, LITERALLY, substracts the mean of each batch (group) from each feature. Perform Batch Mean Centering (BMC) correction.

Parameters:

• data (pandas.DataFrame) – Input data containing OTU counts and metadata.

• sample_label (str) – Column name for sample identifiers.

• batch_label (str) – Column name for batch identifiers.

• exp_label (str) – Column name for experiment/tissue identifiers.

Returns:

DataFrame with sample, experiment, batch, and batch mean centered features.

Return type:

pandas.DataFrame

abaco.batch_correction.correctCombat(data, sample_label='sample', batch_label='batch', experiment_label='tissue')

Perform ComBat batch correction.

Parameters:

• data (pandas.DataFrame) – Input data containing OTU counts and metadata.

• sample_label (str, optional) – Column name for sample identifiers, by default ‘sample’.

• batch_label (str, optional) – Column name for batch identifiers, by default ‘batch’.

• experiment_label (str, optional) – Column name for experiment/tissue identifiers, by default ‘tissue’.

Returns:

DataFrame with sample, batch, experiment, and ComBat-corrected features.

Return type:

pandas.DataFrame

abaco.batch_correction.correctCombatSeq(data, sample_label, batch_label, condition_label, ref_batch=None)

Perform ComBat-seq batch correction for count data.

Parameters:

• data (pandas.DataFrame) – Input data containing count data and metadata.

• sample_label (str) – Column name for sample identifiers.

• batch_label (str) – Column name for batch identifiers.

• condition_label (str) – Column name for condition/experiment identifiers.

• ref_batch (str or None, optional) – Reference batch to use, by default None.

Returns:

DataFrame with sample, batch, condition, and ComBat-seq corrected counts.

Return type:

pandas.DataFrame

abaco.batch_correction.correctConQuR(df, batch_cols, covariate_cols, reference_batch=None, quantiles=(0.05, 0.25, 0.5, 0.75, 0.95), logistic_kwargs={'max_iter': 200, 'penalty': 'l2', 'solver': 'lbfgs'}, quantile_kwargs={'alpha': 0.0})

Conditional logistic quantile regression (ConQuR) for batch correction.

Parameters:

• df (pandas.DataFrame) – Input data containing OTU counts and metadata.

• batch_cols (list of str) – List of batch column names.

• covariate_cols (list of str) – List of covariate column names.

• reference_batch (dict, optional) – Dictionary specifying reference batch values for each batch column. If None, uses zeros for all batch columns.

• quantiles (tuple of float, optional) – Quantiles to use for quantile regression, by default (0.05, 0.25, 0.5, 0.75, 0.95).

• logistic_kwargs (dict, optional) – Keyword arguments for LogisticRegression.

• quantile_kwargs (dict, optional) – Keyword arguments for QuantileRegressor.

Returns:

Batch-corrected DataFrame.

Return type:

pandas.DataFrame

abaco.batch_correction.correctLimma_rBE(data, sample_label='sample', batch_label='batch', covariates_labels=None)

Perform batch correction using Limma’s removeBatchEffect approach.

Parameters:

• data (pandas.DataFrame) – Input data containing OTU counts and metadata.

• sample_label (str, optional) – Column name for sample identifiers, by default ‘sample’.

• batch_label (str, optional) – Column name for batch identifiers, by default ‘batch’.

• covariates_labels (str or list of str, optional) – Additional covariate column(s) to include in the model.

Returns:

DataFrame with original labels and batch-corrected numeric data.

Return type:

pandas.DataFrame

abaco.batch_correction.correctPLSDAbatch(df: DataFrame, sample_label: str, exp_label: str, batch_label: str, ncomp_trt: int = 1, ncomp_batch: int = 1)

Perform PLSDA-batch correction.

Parameters:

• df (pandas.DataFrame) – Input data containing OTU counts and metadata.

• sample_label (str) – Column name for sample identifiers.

• exp_label (str) – Column name for experiment/tissue identifiers.

• batch_label (str) – Column name for batch identifiers.

• ncomp_trt (int, optional) – Number of treatment components, by default 1.

• ncomp_batch (int, optional) – Number of batch components, by default 1.

Returns:

DataFrame with sample, experiment, batch, and PLSDA-batch corrected features.

Return type:

pandas.DataFrame

abaco.batch_correction.correctPLSDAbatch_R(df, sample_label, exp_label, batch_label, ncomp_trt=1, ncomp_bat=1, keepX_trt=None, keepX_bat=None, tol=1e-06, max_iter=500, near_zero_var=True, balance=True)

Python adaptation of PLSDA_batch from R. Returns corrected DataFrame.

Parameters:

• df (pandas.DataFrame) – Input data containing OTU counts and metadata.

• sample_label (str) – Column name for sample identifiers.

• exp_label (str) – Column name for experiment/tissue identifiers.

• batch_label (str) – Column name for batch identifiers.

• ncomp_trt (int, optional) – Number of treatment components, by default 1.

• ncomp_bat (int, optional) – Number of batch components, by default 1.

• keepX_trt (list of int, optional) – Number of variables to keep for each treatment component.

• keepX_bat (list of int, optional) – Number of variables to keep for each batch component.

• tol (float, optional) – Convergence tolerance, by default 1e-6.

• max_iter (int, optional) – Maximum number of iterations, by default 500.

• near_zero_var (bool, optional) – Whether to filter near-zero variance features, by default True.

• balance (bool, optional) – Whether to balance design, by default True.

Returns:

DataFrame with sample, experiment, batch, and corrected features.

Return type:

pandas.DataFrame

abaco.batch_correction.deflate_mtx(X, t)

Deflate matrix X by component t: X - t (t^T t)^{-1} t^T X

Parameters:

• X (numpy.ndarray) – Data matrix to be deflated.

• t (numpy.ndarray) – Component vector.

Returns:

Deflated matrix.

Return type:

numpy.ndarray