Source Code for Module removeIdenticalSeq

 1  #!/env/cns/home/afelten/agc/afelten/script/Python-2.7.5/python 
 2  """#!/usr/bin/python""" 
 3  # -*- coding: iso-8859-1 -*- 
 4  import hashlib 
 5  import argparse 
 6  import os, sys 
 7   
 8   
 9  __doc__=""" 
10  Remove identical sequences from a FASTA file 
11   
12  @requires: U{hashlib<https://docs.python.org/2/library/hashlib.html>} python library 
13   
14  Input : SEQUENCE.fasta (1 line sequence) 
15  Output : PREFIX.fasta 
16  """ 
17   
18 -def get_parser(): 
19           
20          parser = argparse.ArgumentParser(description='Remove identical sequences from a FASTA file') 
21   
22          parser.add_argument('-i', action="store", dest='fasta',  
23                                                  type=str, help='input FASTA file')                       
24                                                   
25          parser.add_argument('-o', action="store", dest='output',  
26                                                  type=str, help='output FASTA file')                                      
27                                                   
28          return parser 
29           
30 -def main():      
31   
32          parser=get_parser()              
33           
34          if len(sys.argv)==1: 
35                  parser.print_help() 
36                  sys.exit(1)      
37           
38          Arguments=parser.parse_args() 
39           
40          fasta = open(Arguments.fasta, "rb") 
41          lignes  = fasta.readlines() 
42          fasta.close() 
43          output = open(Arguments.output, "wb") 
44          md5List = [] 
45           
46          seqNb = 0 
47          outNb = 0 
48           
49          for ligne in lignes: 
50                  if '>' in ligne[0]: # header 
51                          ID = ligne 
52                          seqNb += 1 
53                  else: 
54                          m = hashlib.md5() 
55                          m.update(ligne) 
56                          md5 = m.hexdigest() 
57                          if md5 not in md5List: 
58                                  md5List.append(md5) 
59                                  output.write(ID + ligne) 
60                                  outNb += 1 
61          output.close()   
62          print str(seqNb - outNb) + " sequences removed" 
63           
64  if __name__ == "__main__": 
65          main() 
66