normalization
Overview
This module provides functions for normalizing data, particularly in the context of mass spectrometry. There are two types of normalization including
- Sample normalization - to normalize samples with different total amounts/concentrations.
- Signal normalization - to address the signal drifts in the mass spectrometry data.
Functions
sample_normalization
sample_normalization(feature_table, sample_metadata=None, method='pqn', feature_selection=True)
Normalizes samples using a feature list, typically excluding blank samples.
Parameters:
feature_table(pandas DataFrame): The feature table.sample_metadata(pd.DataFrame): DataFrame containing sample metadata. See params module for details.method(str): The method to find the normalization factors. Options:'pqn': Probabilistic Quotient Normalization.
feature_selection(bool): Whether to select high-quality features for normalization. High-quality features have relative standard deviation (RSD) less than 25% in QC samples and average intensity in QC+biological samples greater than 2 fold of the average intensity in blank samples.
Returns:
pandas DataFrame: The normalized feature table.
find_normalization_factors
find_normalization_factors(array, method='pqn')
Finds the normalization factors for a data frame.
Parameters:
array(numpy array): The data to be normalized.method(str): The method to find the normalization factors. Options:'pqn': Probabilistic Quotient Normalization.
Returns:
numpy array: The normalization factors.
sample_normalization_by_factors
sample_normalization_by_factors(array, v)
Normalizes a data frame by a vector.
Parameters:
array(numpy array): The data to be normalized.v(numpy array): The normalization factor.
Returns:
numpy array: The normalized data.
find_reference_sample
find_reference_sample(array, method='median_intensity')
Finds the reference sample for normalization.
Parameters:
array(numpy array): The data to be normalized.method(str): The method to find the reference sample. Options:'number': The reference sample has the most detected features.'total_intensity': The reference sample has the highest total intensity.'median_intensity': The reference sample has the highest median intensity.
Returns:
int: The index of the reference sample.
high_quality_feature_selection
high_quality_feature_selection(array, is_qc=None, is_blank=None, blank_ratio_tol=0.5, qc_rsd_tol=0.25)
Selects high-quality features based on provided criteria for normalization.
Parameters:
array(numpy array): The data to be normalized. Samples are in columns and features are in rows.is_qc(numpy array): Boolean array indicating whether a sample is a quality control sample.is_blank(numpy array): Boolean array indicating whether a sample is a blank sample.blank_ratio_tol(float): The tolerance of the ratio of the average intensity in blank samples to the average intensity in QC and biological samples.qc_rsd_tol(float): The tolerance of the relative standard deviation (RSD) in QC samples.
Returns:
numpy array: High-quality features. Features are in rows and samples are in columns.numpy array: The index of the selected features.
signal_normalization
signal_normalization(feature_table, sample_metadata, method='lowess', output_plot_path=None)
Normalizes MS signal drifts based on analytical order.
Parameters:
feature_table(pandas DataFrame): The feature table.sample_metadata(pd.DataFrame): DataFrame containing sample metadata. See params module for details.method(str): The method to find the normalization factors. Options:'lowess': Locally Weighted Scatterplot Smoothing.
output_plot_path(str): The path to save the normalization plot. If none, no visualization will be generated.
Returns:
pandas DataFrame: The normalized feature table.
lowess_normalization
lowess_normalization(array, qc_idx, frac=0.07, it=3)
Normalizes samples using quality control samples.
Parameters:
array(numpy array): The data to be normalized.qc_idx(numpy array of bool): Boolean array indicating whether a sample is a quality control sample. It’s length should be the same as the length of array.frac(float): The fraction of the data used when estimating each y-value (used in lowess).it(int): The number of residual-based reweightings to perform (used in lowess).
Returns:
- ‘dict’: A dictionary containing the lowess model, the fit curve, and the normalized array: {‘model’: model, ‘fit_curve’: y, ’normed_arr’: int_arr_corr}