CNNpeaks

#!/usr/bin/python
"""
All annotations of these sources are written with poor english from writer.
Please understand if it will be so crappy. . . .
"""

import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'
import argparse
import sys
import logging

num_grid = 12000
windowSize = 100000

def main():
    global num_grid, windowSize
    #################### Setting arguments ########################
    arg_parser = argparse.ArgumentParser()
    arg_parser.add_argument("-m","--runMode", choices=['preprocess','buildModel','peakCall','checkData','errorCall','randomLabel','inspection'] ,
                            help="Select a mode.")
    arg_parser.add_argument("-i","--inputDir", help="Input directory including labeled data and bam alignment files."
                                                    "\nIn case of callPeak mode, it will be input bam file to call peaks.")
    arg_parser.add_argument("-l","--labelData", help="Label data for error calling.")
    arg_parser.add_argument("-o","--outputBed", default=None, help="Output bedfile name.")
    arg_parser.add_argument("-grid","--gridSize",default=12000, help="Define numbers of grid for each training data")
    arg_parser.add_argument("-s","--searchingDist", help="DBSCAN clustering parameter during preprocessing steps")
    arg_parser.add_argument("-eps","--basePointEPS",help="DBSCAN clustering parameter during preprocessing steps")
    arg_parser.add_argument("-w","--windowSize",default=100000,help="Window size for peak calling.")
    arg_parser.add_argument("-n","--modelName",default=None, help="Define which model to call peak.")
    arg_parser.add_argument("-kf","--KFold",default=10, help="K fold cross validation during training models.")
    arg_parser.add_argument("-ncv","--notCrossValid", action='store_true')
    arg_parser.add_argument("-ln","--labelNum", default=30, help="The number of label for -randomLabel")
    arg_parser.add_argument("-r","--regions", help="Specific genome regions for peak calling. eg) chr3:100023400-1000044323 , you can use 's' or 'e' to indicate start and end of the chromosome")
    arg_parser.add_argument("--broad", help="Broad call mode to call broad histone marks", action='store_true')
    args = arg_parser.parse_args()

    if args.inputDir == None:
        logger.error("'-i' : Input Directory was missed.")
        exit()
    ###############################################################

    num_grid= int(args.gridSize)
    windowSize = int(args.windowSize)
    if args.broad:
        windowSize = int(args.windowSize) * 8 
    
    if args.runMode == 'preprocess':
        from preProcessing.preProcessing import run as preProcessing
        preProcessing(args.inputDir, logger, num_grid=num_grid)
    elif args.runMode =='buildModel':
        from buildModel.buildModel import run as buildModel
        buildModel(args.inputDir, logger, num_grid=num_grid, K_fold_in=int(args.KFold), cross_valid=not (args.notCrossValid))
    elif args.runMode == 'peakCall':
        from peakCalling.callPeaks import run as callPeaks
        callPeaks(args.inputDir, logger, window_size=windowSize, num_grid=num_grid, model_name=args.modelName, regions=args.regions, bed_name=args.outputBed)
    elif args.runMode == 'inspection':
        from inspection.inspection import run as inspection
        
    elif args.runMode == 'checkData':
        from utility.labelManager import labelManager
        labelManager(args.inputDir)
    elif args.runMode == 'errorCall':
        from utility.errorCall import run as errorCall
        errorCall(args.inputDir, args.labelData, logger)
    elif args.runMode == 'randomLabel':
        from utility.randomLabel import run as randomLabel
        randomLabel(args.inputDir)
    else:
        logger.info("-m ( --runMode ) must be one of : { preprocess, buildModel, peakCall , checkData, errorCall }.")



if __name__ == '__main__':
    logging.getLogger('tensorflow').setLevel(logging.FATAL)
    logger = logging.getLogger("ConvLog")
    logger.setLevel(logging.INFO)               # The logger object only output logs which have
                                                # upper level than INFO.
    log_format = logging.Formatter('%(asctime)s:%(message)s')
    stream_handler = logging.StreamHandler()    # Log output setting for the command line.
    stream_handler.setFormatter(log_format)     # The format of stream log will follow this format.
    logger.addHandler(stream_handler)

    #file_handler = logging.FileHandler()        # Log output setting for the file.
    #logger.addHandler(file_handler)

    try:
        main()
    except KeyboardInterrupt:
        sys.stderr.write("USER INTERRUPT. \n")
        sys.exit()