comparison heteroprob.py @ 0:07588b899c13 draft

Uploaded
author arkarachai-fungtammasan
date Wed, 01 Apr 2015 17:05:51 -0400
parents
children
comparison
equal deleted inserted replaced
-1:000000000000 0:07588b899c13
1 ### import libraries ###
2 import sys
3 import collections, math
4 import heapq
5 import itertools
6
7
8
9 ### basic function ###
10 def permuterepeat(n,rlist):
11 f = math.factorial
12 nfac=f(n)
13 rfaclist=[f(i) for i in rlist]
14 for rfac in rfaclist:
15 nfac=nfac/rfac
16 return nfac
17
18 def nCr(n,r):
19 f = math.factorial
20 return f(n) / f(r) / f(n-r)
21
22 def averagelist(a,b,expectedlevelofminor):
23 product=[]
24 for i in range(len(a)):
25 product.append((1-expectedlevelofminor)*a[i]+expectedlevelofminor*b[i])
26
27 return product
28
29 def complement_base(read):
30 collect=''
31 for i in read:
32 if i.upper()=='A':
33 collect+='T'
34 elif i.upper()=='T':
35 collect+='A'
36 elif i.upper()=='C':
37 collect+='G'
38 elif i.upper()=='G':
39 collect+='C'
40 return collect
41 def makeallpossible(read):
42 collect=[]
43 for i in range(len(read)):
44 tmp= read[i:]+read[:i]
45 collect.append(tmp)
46 collect.append(complement_base(tmp))
47 return collect
48
49 def motifsimplify(base):
50 '''str--> str
51 '''
52 motiflength=len(base)
53 temp=list(set(ALLMOTIF[motiflength]).intersection(set(makeallpossible(base))))
54
55 return temp[0]
56
57 def majorallele(seq):
58 binseq=list(set(seq))
59 binseq.sort(reverse=True) # highly mutate mode
60 #binseq.sort() # majority mode
61 storeform=''
62 storevalue=0
63 for i in binseq:
64 if seq.count(i)>storevalue:
65 storeform=i
66 storevalue=seq.count(i)
67
68 return int(storeform)
69
70 ### decide global parameter ###
71 COORDINATECOLUMN=1
72 ALLELECOLUMN=2
73 MOTIFCOLUMN=3
74 inputname=sys.argv[1]
75 errorprofile=sys.argv[2]
76 EXPECTEDLEVELOFMINOR=float(sys.argv[3])
77 if EXPECTEDLEVELOFMINOR >0.5:
78 try:
79 errorexpectcontribution=int('a')
80 except Exception, eee:
81 print eee
82 stop_err("Expected contribution of minor allele must be at least 0 and not more than 0.5")
83 MINIMUMMUTABLE=0 ###1.2*(1.0/(10**8)) #http://www.ncbi.nlm.nih.gov/pubmed/22914163 Kong et al 2012
84
85
86 ## Fixed global variable
87 ALLREPEATTYPE=[1,2,3,4]
88 ALLREPEATTYPENAME=['mono','di','tri','tetra']
89 monomotif=['A','C']
90 dimotif=['AC','AG','AT','CG']
91 trimotif=['AAC','AAG','AAT','ACC','ACG','ACT','AGC','AGG','ATC','CCG']
92 tetramotif=['AAAC','AAAG','AAAT','AACC','AACG','AACT','AAGC','AAGG','AAGT','AATC','AATG','AATT',\
93 'ACAG','ACAT','ACCC','ACCG','ACCT','ACGC','ACGG','ACGT','ACTC','ACTG','AGAT','AGCC','AGCG','AGCT',\
94 'AGGC','AGGG','ATCC','ATCG','ATGC','CCCG','CCGG','AGTC']
95 ALLMOTIF={1:monomotif,2:dimotif,3:trimotif,4:tetramotif}
96 monorange=range(5,60)
97 dirange=range(6,60)
98 trirange=range(9,60)
99 tetrarange=range(12,80)
100 ALLRANGE={1:monorange,2:dirange,3:trirange,4:tetrarange}
101
102 #########################################
103 ######## Prob calculation sector ########
104 #########################################
105 def multinomial_prob(majorallele,STRlength,motif,probdatabase):
106 '''int,int,str,dict-->int
107 ### get prob for each STRlength to be generated from major allele
108 '''
109 #print (majorallele,STRlength,motif)
110 prob=probdatabase[len(motif)][motif][majorallele][STRlength]
111 return prob
112
113 ################################################
114 ######## error model database sector ###########
115 ################################################
116
117 ## structure generator
118 errormodeldatabase={1:{},2:{},3:{},4:{}}
119 sumbymajoralleledatabase={1:{},2:{},3:{},4:{}}
120 for repeattype in ALLREPEATTYPE:
121 for motif in ALLMOTIF[repeattype]:
122 errormodeldatabase[repeattype][motif]={}
123 sumbymajoralleledatabase[repeattype][motif]={}
124 for motifsize1 in ALLRANGE[repeattype]:
125 errormodeldatabase[repeattype][motif][motifsize1]={}
126 sumbymajoralleledatabase[repeattype][motif][motifsize1]=0
127 for motifsize2 in ALLRANGE[repeattype]:
128 errormodeldatabase[repeattype][motif][motifsize1][motifsize2]=MINIMUMMUTABLE
129 #print errormodeldatabase
130 ## read database
131
132 ## get read count for each major allele
133 fd=open(errorprofile)
134 lines=fd.readlines()
135 for line in lines:
136 temp=line.strip().split('\t')
137 t_major=int(temp[0])
138 t_count=int(temp[2])
139 motif=temp[3]
140 sumbymajoralleledatabase[len(motif)][motif][t_major]+=t_count
141 fd.close()
142 ##print sumbymajoralleledatabase
143
144 ## get probability
145 fd=open(errorprofile)
146 lines=fd.readlines()
147 for line in lines:
148 temp=line.strip().split('\t')
149 t_major=int(temp[0])
150 t_read=int(temp[1])
151 t_count=int(temp[2])
152 motif=temp[3]
153 if sumbymajoralleledatabase[len(motif)][motif][t_major]>0:
154 errormodeldatabase[len(motif)][motif][t_major][t_read]=t_count/(sumbymajoralleledatabase[len(motif)][motif][t_major]*1.0)
155 #errormodeldatabase[repeattype][motif][t_major][t_read]=math.log(t_count/(sumbymajorallele[t_major]*1.0))
156
157 #else:
158 # errormodeldatabase[repeattype][motif][t_major][t_read]=0
159 fd.close()
160 #print errormodeldatabase
161 #print math.log(100,10)
162 #########################################
163 ######## input reading sector ###########
164 #########################################
165
166
167
168 fd = open(inputname)
169 ##fd=open('sampleinput_C.txt')
170 lines=fd.xreadlines()
171 for line in lines:
172 i_read=[]
173 i2_read=[]
174 temp=line.strip().split('\t')
175 i_coordinate=temp[COORDINATECOLUMN-1]
176 i_motif=motifsimplify(temp[MOTIFCOLUMN-1])
177 i_read=temp[ALLELECOLUMN-1].split(',')
178 i_read=map(int,i_read)
179 depth=len(i_read)
180 heteromajor1=int(temp[6])
181 heteromajor2=int(temp[7])
182
183 ### calculate the change to detect combination (using error profile)
184 heterozygous_collector=0
185 alist=[multinomial_prob(heteromajor1,x,i_motif,errormodeldatabase)for x in i_read]
186 blist=[multinomial_prob(heteromajor2,x,i_motif,errormodeldatabase)for x in i_read]
187
188 ablist=averagelist(alist,blist,EXPECTEDLEVELOFMINOR)
189
190 if 0 in ablist:
191 continue
192 heterozygous_collector=reduce(lambda y, z: y*z,ablist )
193
194 ### prob of combination (using multinomial distribution)
195 frequency_distribution=[len(list(group)) for key, group in itertools.groupby(i_read)]
196 ## print frequency_distribution
197 expandbypermutation=permuterepeat(depth,frequency_distribution)
198
199 print line.strip()+'\t'+str(heterozygous_collector)+'\t'+str(expandbypermutation)+'\t'+str(expandbypermutation*heterozygous_collector)+'\t'+str(depth)