MakeBootstraps.py

#!/usr/bin/env python
'''
@author: jonathanfriedman
Script for making simulated datasets used to get pseudo p-values.

'''
import os
import numpy as np 
import pandas as pd 
from sparcc.io_methods import read_txt, write_txt
from typing import Union

def permute_w_replacement(frame:Union[pd.DataFrame,np.ndarray], axis=0):
    '''
    Permute the frame values across the given axis.
    Create simulated dataset were the counts of each component (column)
    in each sample (row), are randomly sampled from the all the 
    counts of that component in all samples.
    
    Parameters
    ----------
    frame : Numpy Array
        Frame to permute.
    axis : {0, 1}
        - 0 - Permute row values across columns
        - 1 - Permute column values across rows    
    
    Returns
    -------
    Permuted DataFrame (new instance).
    '''
 
    if isinstance(frame,pd.DataFrame):
        frame=frame.values

    fp=lambda x:np.random.permutation(x)  

    if axis==0:
        #Along the columns
        aux=np.apply_along_axis(fp,0,frame)        
        aux=np.apply_along_axis(fp,1,aux)
        return  aux

    elif axis==1:
        #Along the rows
        aux=np.apply_along_axis(fp,1,frame)
        aux=np.apply_along_axis(fp,0,aux)
        return aux
    

def make_bootstraps(counts, nperm, perm_template, outpath='./', iprint=0):
    '''
    Make n simulated datasets used to get pseudo p-values. 
    Simulated datasets are generated by assigning each OTU in each sample 
    an abundance that is randomly drawn (w. replacement) from the 
    abundances of the OTU in all samples. 
    Simulated datasets are either written out as txt files.

    Parameters
    ----------
    counts : DataFrame
        Inferred correlations whose p-values are to be computed.
    nperm : int
        Number of permutations to produce.
    perm_template : str
        Template for the permuted data file names.
        Should not include the path, which is specified using the 
        outpath parameter.
        The iteration number is indicated with a "#".
        For example: 'permuted/counts.permuted_#.txt'
    outpath : str (default './')
        The path to which permuted data will be written.
        If not provided files will be written to the cwd.
    iprint : int (default = 0)
        The interval at which iteration number is printed out.
        If iprint<=0 no printouts are made.
    '''
    if not os.path.exists(outpath): os.makedirs(outpath)

    for i in range(nperm): 
        if iprint>0:
            if not i%iprint: print(i)
        #New Matrix
        counts_perm = permute_w_replacement(counts, axis=1) 

        outfile = outpath + perm_template.replace('#', '%d'%i)
        #The output is written
        write_txt(counts_perm, outfile,index=True)

def main(counts_file, nperm, perm_template, outpath='./'):
    '''
    Make n simulated datasets used to get pseudo p-values. 
    Simulated datasets are generated by assigning each OTU in each sample 
    an abundance that is randomly drawn (w. replacement) from the 
    abundances of the OTU in all samples. 
    Simulated datasets are either written out as csv files.
    '''
    if perm_template is None:
        perm_template = counts_file + '.permuted_#.csv'
    ## read counts data
    counts = read_txt(counts_file,index_col=0)
    if counts.shape[0]==0:
        print('A problem has occurred with the file, it will be resolved.')
        
        try:
            counts=read_txt(counts_file,sep=',',index_col=0)
        except IOError as IOE:
            raise (IOE)
    assert counts.shape[0]!=0,"ERROR!"

    ## make permutated data
    print("Permutations")
    make_bootstraps(counts, nperm, perm_template, outpath=outpath)

if __name__ == '__main__':

    ## parse input arguments
    from optparse import OptionParser
    kwargs = {}
    usage  = ('Make n simulated datasets used to get pseudo p-values.\n' 
              'Simulated datasets are generated by assigning each OTU in each sample an abundance that is randomly drawn (w. replacement) from the abundances of the OTU in all samples.\n' 
              'Simulated datasets are either written out as txt files. \n'
              '\n'
              'Usage:   python MakeBootstraps.py counts_file [options]\n'
              'Example: python MakeBootstraps.py example/fake_data.txt -n 5 -t permutation_#.txt -p example/pvals/')
    parser = OptionParser(usage)
    parser.add_option("-n", dest="n", default=100, type = 'int',
                      help="Number of simulated datasets to create (100 default).")
    parser.add_option("-t", "--template", dest="perm_template", default=None, type = 'str',
                      help="The template for the permuted data file names.\n"
                           "Should not include the path, which is specified using the -p option.\n"
                           'The iteration number is indicated with a "#".\n'
                           "For example: 'permuted/counts.permuted_#.txt'" 
                           "If not provided a '.permuted_#.txt' suffix will be added to the counts file name.\n")
    parser.add_option("-p", "--path", dest="outpath", default='./', type = 'str',
                      help="The path to which permuted data will be written.\n" 
                           "If not provided files will be written to the cwd.\n")
    (options, args) = parser.parse_args()
    counts_file     = args[0]
    n               = options.n
    outpath         = options.outpath
    perm_template   = options.perm_template
    
    main(counts_file, n, perm_template, outpath)