Source code for Protomix.binning

import numpy as np
import pandas as pd



def binning(spectra_df: pd.DataFrame, bin_size: float, method="trapezoidal") -> pd.DataFrame:
    """
    Perform uniform binning on an NMR spectrum by integrating the area under the curve for each bin.

    This function applies uniform binning to the provided NMR spectra, dividing the spectrum into bins of 
    a specified size and integrating the area under the curve within each bin. The integration can be performed 
    using either the trapezoidal or rectangular method.

    :param spectra_df: A DataFrame containing the spectra, where each row represents a sample, and columns represent ppm values.
    :type spectra_df: pd.DataFrame
    :param bin_size: The size of each bin in ppm.
    :type bin_size: float
    :param method: The integration method to use for binning. Options are "trapezoidal" (default) or "rectangular".
    :type method: str
    
    :return: A DataFrame containing the binned spectra with integrated intensities for each bin.
    :rtype: pd.DataFrame
    """
    
    # Extract ppm values from columns
    ppm_values = spectra_df.columns.astype(float).values
    
    # Define the start and end of the ppm range
    ppm_start = ppm_values[0]
    ppm_end = ppm_values[-1]
    
    # Create new binned ppm values
    binned_ppm = np.arange(ppm_start, ppm_end, bin_size)
    
    # Create an empty DataFrame to store binned data
    binned_df = pd.DataFrame(index=spectra_df.index, columns=binned_ppm[:-1])
    
    for i in range(len(binned_ppm) - 1):
        # Mask to select values in the current bin range
        mask = (ppm_values >= binned_ppm[i]) & (ppm_values < binned_ppm[i+1])
        
        # Determine the width of each data point within the bin
        delta_ppm = ppm_values[1] - ppm_values[0]
        
        if method == "trapezoidal":
            # Average the first and last intensities in the bin for each sample
            avg_intensity = 0.5 * (spectra_df.loc[:, mask].iloc[:, 0] + spectra_df.loc[:, mask].iloc[:, -1])
            binned_df[binned_ppm[i]] = avg_intensity * delta_ppm * mask.sum()
        
        elif method == "rectangular":
            # Use the midpoint intensity for each bin (or the first intensity, depending on preference)
            binned_df[binned_ppm[i]] = spectra_df.loc[:, mask].mean(axis=1) * delta_ppm * mask.sum()
        else:
            raise ValueError("Invalid method. Choose 'trapezoidal' or 'rectangular'.")
    
    return binned_df