Mercurial > repos > xuebing > sharplab_seq_motif

comparison match.cpp @ 15:0e221dbd17b2 default tip

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Uploaded
author xuebing
date Sat, 31 Mar 2012 08:53:06 -0400
parents
children
comparison
equal deleted inserted replaced
14:d1f0f85ee5bc 15:0e221dbd17b2
1 #include <string>
2 #include <sstream>
3 #include <iostream>
4 #include <fstream>
5 #include <vector>
6 #include <set>
7 #include <map>
8
9 #include <string.h>
10 #include <stdio.h>
11 #include <stdlib.h>
12 #include <time.h>
13
14 using namespace std;
15
16 string current_time()
17 {// get current time, in the format:
18 // Thu Mar 15 21:06:57 2012
19 time_t rawtime;
20 struct tm * timeinfo;
21 time ( &rawtime );
22 timeinfo = localtime ( &rawtime );
23 string str = asctime (timeinfo);
24 return str.substr(0,str.size()-1);
25 }
26
27 template<class key,class val>
28 void PrintMap(map<key,val> m)
29 {// print out a map
30 for (typename map<key,val>::iterator it = m.begin();it!=m.end(); it++)
31 {
32 cout << (*it).first << "\t=>\t" <<(*it).second<<endl;
33 }
34 }
35
36 char complement(char ch){
37 switch (ch)
38 {
39 case 'A':return 'T';
40 case 'C':return 'G';
41 case 'G':return 'C';
42 case 'T':return 'A';
43 default:return 'N';
44 }
45 }
46
47 string reverseComplement(string seq){
48 int L = seq.length();
49 string rc (L,'0');
50 for (int i=0;i<L; i++)
51 {
52 rc[L-i-1] = complement(seq[i]);
53 }
54 return rc;
55 }
56
57 void ReadOneSeqFromFasta(ifstream& infile, string& name, string& seq)
58 {
59 // read one sequence from fasta file
60 getline(infile,name); // read identifier line
61 name = name.substr(1);// remove leading '>'
62 seq = ""; // initialize sequence
63 string str;
64 while(infile.peek() != '>' && infile.good())
65 {// before next '>' and before hitting the end of the file
66 getline(infile,str);
67 seq.append(str);
68 }
69 }
70
71 vector<int> findall(string seq, string motif)
72 {
73 // find all match positions of motif in seq
74 vector<int> allpos;
75 size_t pos = 0;
76 while(pos != string::npos)
77 {
78 pos = seq.find(motif,pos+1);
79 allpos.push_back(pos);
80 }
81 return allpos;
82 }
83
84 string int2str(int number)
85 {
86 stringstream ss;//create a stringstream
87 ss << number;//add number to the stream
88 return ss.str();//return a string with the contents of the stream
89 }
90
91
92
93 map<string,string> ReadFasta(string filename)
94 {//read all sequences in a fasta file
95 ifstream fin(filename.c_str());
96
97 map<string,string> seqs;
98 string name,seq;
99 while(fin.good())
100 {
101 ReadOneSeqFromFasta(fin,name,seq);
102 seqs[name] = seq;
103 }
104 fin.close();
105 return seqs;
106 }
107
108 void mismatches(map<string,string>& mutant,map<string,int>& dist, string motif, int n, set<char> alphabet)
109 {
110 set<char>::iterator it;
111 if (mutant.count(motif) == 0)
112 {
113 mutant[motif] = "";
114 dist[motif]=n;
115 }
116 if(n==0){return;}
117 for (int i=0;i<motif.length();i++)
118 {
119 string str=motif;
120 set<char> ab = alphabet;
121 ab.erase(str[i]);
122 for (it = ab.begin(); it!=ab.end(); it++)
123 {
124 str[i] = *it;
125 //cout << "mutate "<<motif<<" to "<<str<<endl;
126 if (mutant.count(str) >0)
127 {
128 if(dist[str] >= n)
129 {
130 //cout << mutant[str] <<endl;
131 continue;
132 }
133 }
134
135 //mutated to a new sequence
136 //cout <<"new mutation"<<endl;
137 mutant[str] = mutant[motif];
138 mutant[str].push_back(',');
139 mutant[str].push_back(motif[i]);
140 mutant[str].append(int2str(i+1));
141 mutant[str].push_back(str[i]);
142 dist[str]=n;
143 //cout << "tag="<<mutant[str]<<" dist="<<n<<endl;
144
145 if (n>1)
146 {
147 //cout << "subproc" <<endl;
148 mismatches(mutant,dist,str,n-1,alphabet);
149 }
150 }
151
152 }
153 }
154
155 map<string,string> ExpandMotifs(map<string,string>& motifs, int nmismatch, bool rc, set<char> alphabet)
156 {
157 map<string,string> expandedmotifs;
158 // generate mismatched motifs
159 map<string,string> mutants;
160 map<string,int> dist;
161 map<string,string>::iterator it; // iterator for motifs
162 map<string,int>::iterator it2; // iterator for mutants
163 string name,seq,tmp;
164 //cout<<"input motifs"<<endl;
165 //PrintMap(motifs);
166 for(it=motifs.begin();it!=motifs.end();it++)
167 {
168 name = (*it).first;
169 seq = (*it).second;
170
171 mismatches(mutants,dist,seq,nmismatch,alphabet);
172 //cout << mutants.size()<<" mutants identified" <<endl;
173 //PrintMap(mutants);
174 // add mutants to motifs
175 for(it2=dist.begin();it2!=dist.end();it2++)
176 {
177 string tmp = name;
178 tmp.append(",").append((*it2).first);
179 tmp.append(mutants[(*it2).first]);
180 expandedmotifs[tmp] = (*it2).first;
181 //cout << name <<","<<tmp<<","<<expandedmotifs[tmp]<<endl;
182 }
183 // clear the mutants list
184 mutants.clear();
185 dist.clear();
186 }
187 //PrintMap(expandedmotifs);
188 //cout << expandedmotifs.size() <<" expanded motifs"<<endl;
189 //cout <<"add reverse complement"<<endl;
190 map<string,string> expandedmotifs_rc = expandedmotifs;
191 if (rc)
192 {
193 for(it=expandedmotifs.begin();it!=expandedmotifs.end();it++)
194 {
195 name = (*it).first;
196 expandedmotifs_rc[name.append(",rc")] = reverseComplement((*it).second);
197 }
198 }
199
200 return expandedmotifs_rc;
201
202 }
203
204
205
206 int* match(string motiffile, string seqfile, string outfile, int nmismatch, bool rc, set<char> alphabet)
207 {
208 int nsite = 0; // total number of matches to report
209 int nseq = 0;
210 ifstream fmotif, fseq;
211 ofstream fout;
212
213 // load motifs
214 map<string,string> motifs = ReadFasta(motiffile);
215 cout <<"["<<current_time()<<"] "<<motifs.size()<< " motifs loaded from "<<motiffile<<endl;
216
217 // expand motifs
218 map<string,string> expandedmotifs = ExpandMotifs(motifs,nmismatch,rc,alphabet);
219 cout <<"["<<current_time()<<"] "<<expandedmotifs.size()<< " motifs after expanding (mismatch/reverse-complement)"<<endl;
220
221 //PrintMap(expandedmotifs);
222
223 // searching motifs in each sequence
224 fseq.open(seqfile.c_str());
225 fout.open(outfile.c_str());
226
227 string seqname,seq,motifname,motif;
228 while(fseq.good())
229 {
230 // read one sequence
231 ReadOneSeqFromFasta(fseq,seqname,seq);
232 nseq = nseq + 1;
233
234 cout.flush();
235 // iterate over motifs
236 map<string,string>::iterator it;
237 for(it=expandedmotifs.begin();it!=expandedmotifs.end();it++)
238 {
239 motifname = (*it).first;
240 motif = (*it).second;
241 //cout << "searching for "<<motifname<<":"<< motif <<endl;
242 vector<int> found = findall(seq,motif);
243 for (int i =0;i<found.size()-1;i++)
244 {
245 fout <<seqname<<"\t"<<found[i]<< "\t"<< motifname <<"\t"<<motif<<endl;
246 }
247 nsite = nsite + found.size()-1;
248 }
249 cout <<"\r["<<current_time()<<"] " << nsite << " sites found in "<< nseq << " sequences " ;
250 }
251
252 cout << endl;
253 fseq.close();
254 fout.close();
255
256 int* res = new int[2];
257 res[0] = nsite;
258 res[1] = nseq;
259
260 return res;
261 }
262
263 vector<string> StringExplode(string str, string separator){
264 int found;
265 vector<string> results;
266 found = str.find_first_of(separator);
267 while(found != string::npos){
268 if(found > 0){
269 results.push_back(str.substr(0,found));
270 }
271 str = str.substr(found+1);
272 found = str.find_first_of(separator);
273 }
274 if(str.length() > 0){
275 results.push_back(str);
276 }
277 return results;
278 }
279
280 int tab2bed(string infile, string outfile)
281 {
282 //hg18_chr6_122208322_122209078_+ 635 5ss,A7C,G8T-rc AAGTACCTG
283 //hg18_chr6_122208322_122209078_+ 553 5ss,C1G,G3A GAAGTAAGT
284 ifstream fin;
285 ofstream fout;
286 fin.open(infile.c_str());
287 fout.open(outfile.c_str());
288 string line;
289 vector<string> flds;
290 vector<string> pos;
291 vector<string> nm;
292 while(fin)
293 {
294 getline(fin,line);
295 if (line.length() == 0)
296 continue;
297 flds = StringExplode(line,"\t");
298 pos = StringExplode(flds[0],"_");
299 if (pos.size() < 5)
300 {
301 cout << "\n!! incorrect sequence name format!\n make sure sequence name looks like: hg18_chr6_122208322_122209078_+\n\n";
302 return 0;
303 }
304 string chr = pos[1];
305 int start = atoi(pos[2].c_str());
306 int end = atoi(pos[3].c_str());
307 int match_start = atoi(flds[1].c_str());
308 int motifLen = flds[3].length();
309 // check if match on the other strand
310 string strandness = "sense";
311 if (flds[2].find("rc") !=string::npos)
312 strandness = "antisense";
313 string strand = pos[4];
314
315 if (strand== "+")
316 {
317 start = start + match_start;
318 if (strandness == "antisense")
319 {
320 strand = "-";
321 }
322 }
323 else//sequence on the - strand of the genome
324 {
325 start = end - match_start - motifLen;
326 if (strandness == "antisense")
327 {
328 strand = "+";
329 }
330 }
331 end = start + motifLen;
332 // number of mismatches
333 nm = StringExplode(flds[2],",");
334 int score = nm.size()-1;
335
336 fout << chr <<"\t"<<start<<"\t" <<end<<"\t"<<flds[3]<< "\t"<<score <<"\t"<<strand<< "\t"<<strandness<<"\t"<<flds[2]<<"\t"<<flds[0]<<"\t"<<flds[1]<<endl;
337
338 }
339 fin.close();
340 fout.close();
341 return 1;//return 1 if successfully created bed file
342 }
343
344 int main(int argc, char* argv[]) {
345
346 if (argc < 7)
347 {
348 cout << "usage\n";
349 cout << " match motif.fa seq.fa outfile nmismatch rc/norc bed/nobed" << endl;
350 cout << "==========\n";
351 cout << "arguments:" <<endl;
352 cout << "==========\n";
353 cout << " 1. motif file, fasta format, could include multiple sequences\n";
354 cout << " 2. sequence file, fasta format, could include multiple sequences\n";
355 cout << " 3. output file name\n";
356 cout << " 4. to search on reverse complement strand or not. use 'rc' or 'norc'\n";
357 cout << " 5. to output BED format or not. use 'bed' only if sequence have id like: hg18_chr12_52642148_52644699_+\n";
358 cout << "==========\n";
359 cout << "output format\n";
360 cout << "==========\n";
361 cout << " tabular:\n";
362 cout << "==========\n";
363 cout << " hg18_chr10_20048197_20048851_+ 21 5ss,T5A:CAGGAAAGT-rc ACTTTCCTG\n";
364 cout << " column 1: sequence id\n";
365 cout << " column 2: start of the match in sequence\n";
366 cout << " column 3: motif variant id, format: motif_id,mismatch:motif_sequence[-reverse_complement]\n";
367 cout << " column 4: actual match in target sequence\n";
368 cout << "==========\n";
369 cout << " BED format:\n";
370 cout << "==========\n";
371 cout << " column 1-3: genomic coordinates of the match, calculated using sequence ID: hg18_chr12_52642148_52644699_+ and match start in the sequence\n";
372 cout << " column 4: number of mismatches\n";
373 cout << " column 6: strand of the match sequence\n";
374 cout << " column 7: strandness of the match relative to the target sequence\n";
375 return 1;
376 }
377
378
379 string motiffile = argv[1]; // motif file
380 string seqfile = argv[2]; // sequence file
381 string outfile = argv[3]; // output file
382 int nmismatch = atoi(argv[4]);// # of mismatches allowed
383 string _rc = argv[5]; // rc or norc: to search reverse complement strand or not
384 string _bed = argv[6]; // to make bed output or not. only valid if sequence have id like: hg18_chr12_52642148_52644699_+
385 bool rc = false;
386 if (_rc == "rc") {rc=true;}
387 bool bed = false;
388 if (_bed == "bed") {bed=true;}
389
390 // alphabet
391 string ACGT_str = "ACGT";
392 set<char> alphabet;
393 for (int i=0;i<4;i++)
394 {
395 alphabet.insert(ACGT_str[i]);
396 }
397
398 int *res = match(motiffile, seqfile, outfile, nmismatch, rc, alphabet);
399
400 if (_bed == "bed" && res[0]>0)
401 {
402 cout <<"["<<current_time()<<"] creating BED format output..."<<endl;
403 string bedfile = outfile;
404 bedfile += ".bed";
405 if (tab2bed(outfile,bedfile))
406 {
407 rename(bedfile.c_str(),outfile.c_str());
408 }
409 }
410 cout <<"["<<current_time()<<"] Done"<<endl;
411 return 0;
412 }
413