1 // This file is part of the Mussa source distribution.
2 // http://mussa.caltech.edu/
3 // Contact author: Tristan De Buysscher, tristan@caltech.edu
5 // This program and all associated source code files are Copyright (C) 2005
6 // the California Institute of Technology, Pasadena, CA, 91125 USA. It is
7 // under the GNU Public License; please see the included LICENSE.txt
8 // file for more information, or contact Tristan directly.
11 // This file is part of the Mussa source distribution.
12 // http://mussa.caltech.edu/
13 // Contact author: Tristan De Buysscher, tristan@caltech.edu
15 // This program and all associated source code files are Copyright (C) 2005
16 // the California Institute of Technology, Pasadena, CA, 91125 USA. It is
17 // under the GNU Public License; please see the included LICENSE.txt
18 // file for more information, or contact Tristan directly.
21 // ----------------------------------------
22 // ---------- sequence.cc -----------
23 // ----------------------------------------
24 #include <boost/filesystem/fstream.hpp>
25 #include <boost/filesystem/operations.hpp>
26 namespace fs = boost::filesystem;
28 #include <boost/spirit/core.hpp>
29 #include <boost/spirit/actor/push_back_actor.hpp>
30 #include <boost/spirit/iterator/file_iterator.hpp>
31 #include <boost/spirit/utility/chset.hpp>
32 namespace spirit = boost::spirit;
34 #include "alg/sequence.hpp"
35 #include "mussa_exceptions.hpp"
51 annot::annot(int begin, int end, std::string type, std::string name)
63 bool operator==(const annot& left, const annot& right)
65 return ((left.begin== right.begin) and
66 (left.end == right.end) and
67 (left.type == right.type) and
68 (left.name == right.name));
71 motif::motif(int begin, std::string motif)
72 : annot(begin, begin+motif.size(), "motif", motif),
82 Sequence::Sequence(AlphabetRef alphabet)
83 : seq(new SeqSpan("", alphabet, SeqSpan::PlusStrand))
91 Sequence::Sequence(const char *seq, AlphabetRef alphabet_, SeqSpan::strand_type strand_)
95 set_filtered_sequence(seq, alphabet_, 0, npos, strand_);
98 Sequence::Sequence(const std::string& seq,
99 AlphabetRef alphabet_,
100 SeqSpan::strand_type strand_)
104 set_filtered_sequence(seq, alphabet_, 0, seq.size(), strand_);
107 Sequence::Sequence(const Sequence& o)
112 motif_list(o.motif_list)
116 Sequence::Sequence(const Sequence* o)
121 motif_list(o->motif_list)
125 Sequence::Sequence(const SeqSpanRef& seq_ref)
132 Sequence &Sequence::operator=(const Sequence& s)
139 motif_list = s.motif_list;
144 static void multiplatform_getline(std::istream& in, std::string& line)
149 while(in.good() and !(c == '\012' or c == '\015') ) {
153 // if we have cr-lf eat it
155 if (c=='\012' or c == '\015') {
160 void Sequence::load_fasta(fs::path file_path, int seq_num, int start_index, int end_index)
162 load_fasta(file_path, reduced_nucleic_alphabet, seq_num, start_index, end_index);
165 //! load a fasta file into a sequence
166 void Sequence::load_fasta(fs::path file_path, AlphabetRef a,
167 int seq_num, int start_index, int end_index)
169 fs::fstream data_file;
170 data_file.open(file_path, std::ios::in);
172 if (!data_file.good())
174 throw mussa_load_error("Sequence File: "+file_path.string()+" not found");
177 load_fasta(data_file, a, seq_num, start_index, end_index);
178 } catch(sequence_empty_error e) {
179 // there doesn't appear to be any sequence
180 // catch and rethrow to include the filename
181 std::stringstream msg;
182 msg << "The selected sequence in "
183 << file_path.native_file_string()
184 << " appears to be empty";
185 throw sequence_empty_error(msg.str());
186 } catch(sequence_empty_file_error e) {
187 std::stringstream errormsg;
188 errormsg << file_path.native_file_string()
189 << " did not have any fasta sequences" << std::endl;
190 throw sequence_empty_file_error(errormsg.str());
191 } catch(sequence_invalid_load_error e) {
192 std::ostringstream msg;
193 msg << file_path.native_file_string();
194 msg << " " << e.what();
195 throw sequence_invalid_load_error(msg.str());
200 void Sequence::load_fasta(std::istream& file,
201 int seq_num, int start_index, int end_index)
203 load_fasta(file, reduced_nucleic_alphabet, seq_num, start_index, end_index);
207 Sequence::load_fasta(std::istream& data_file, AlphabetRef a,
209 int start_index, int end_index)
211 std::string file_data_line;
212 int header_counter = 0;
213 size_t line_counter = 0;
214 bool read_seq = true;
215 std::string rev_comp;
216 std::string sequence_raw;
217 std::string seq_tmp; // holds sequence during basic filtering
218 const Alphabet &alpha = Alphabet::get_alphabet(a);
221 throw mussa_load_error("fasta sequence number is 1 based (can't be 0)");
224 // search for the header of the fasta sequence we want
225 while ( (!data_file.eof()) && (header_counter < seq_num) )
227 multiplatform_getline(data_file, file_data_line);
229 if (file_data_line.substr(0,1) == ">")
233 if (header_counter > 0) {
234 header = file_data_line.substr(1);
238 while ( !data_file.eof() && read_seq ) {
239 multiplatform_getline(data_file,file_data_line);
241 if (file_data_line.substr(0,1) == ">")
244 for (std::string::const_iterator line_i = file_data_line.begin();
245 line_i != file_data_line.end();
248 if(alpha.exists(*line_i)) {
249 sequence_raw += *line_i;
251 std::ostringstream msg;
252 msg << "Unrecognized characters in fasta sequence at line ";
254 throw sequence_invalid_load_error(msg.str());
260 // Lastly, if subselection of the sequence was specified we keep cut out
261 // and only keep that part
262 // end_index = 0 means no end was specified, so cut to the end
264 end_index = sequence_raw.size();
266 // sequence filtering for upcasing agctn and convert non AGCTN to N
267 if (end_index-start_index <= 0) {
268 std::string msg("The selected sequence appears to be empty");
269 throw sequence_empty_error(msg);
271 set_filtered_sequence(sequence_raw, a, start_index, end_index-start_index, SeqSpan::PlusStrand);
273 std::string errormsg("There were no fasta sequences");
274 throw sequence_empty_file_error(errormsg);
278 void Sequence::set_filtered_sequence(const std::string &in_seq,
279 AlphabetRef alphabet_,
282 SeqSpan::strand_type strand_)
285 count = in_seq.size() - start;
287 new_seq.reserve(count);
289 // finally, the actual conversion loop
290 const Alphabet& alpha_impl = Alphabet::get_alphabet(alphabet_); // go get one of our actual alphabets
291 std::string::const_iterator seq_i = in_seq.begin()+start;
292 for(size_type i = 0; i != count; ++i, ++seq_i)
294 if (alpha_impl.exists(*seq_i)) {
295 new_seq.append(1, toupper(*seq_i));
297 new_seq.append(1, 'N');
300 SeqSpanRef new_seq_ref(new SeqSpan(new_seq, alphabet_, strand_));
305 Sequence::load_annot(fs::path file_path, int start_index, int end_index)
307 if (not fs::exists(file_path)) {
308 throw mussa_load_error("Annotation File " + file_path.string() + " was not found");
310 if (fs::is_directory(file_path)) {
311 throw mussa_load_error(file_path.string() +
312 " is a directory, please provide a file for annotations."
315 fs::fstream data_stream(file_path, std::ios::in);
318 throw mussa_load_error("Error loading annotation file " + file_path.string());
321 // so i should probably be passing the parse function some iterators
322 // but the annotations files are (currently) small, so i think i can
323 // get away with loading the whole file into memory
326 while(data_stream.good()) {
333 parse_annot(data, start_index, end_index);
334 } catch(annotation_load_error e) {
335 std::ostringstream msg;
336 msg << file_path.native_file_string()
339 throw annotation_load_error(msg.str());
343 /* If this works, yikes, this is some brain hurting code.
345 * what's going on is that when pb_annot is instantiated it stores references
346 * to begin, end, name, type, declared in the parse function, then
347 * when operator() is called it grabs values from those references
348 * and uses that to instantiate an annot object and append that to our
351 * This weirdness is because the spirit library requires that actions
352 * conform to a specific prototype operator()(IteratorT, IteratorT)
353 * which doesn't provide any useful opportunity for me to actually
354 * grab the results of our parsing.
356 * so I instantiate this structure in order to have a place to grab
360 struct push_back_annot {
361 std::list<annot>& annot_list;
368 push_back_annot(std::list<annot>& annot_list_,
374 : annot_list(annot_list_),
383 void operator()(std::string::const_iterator,
384 std::string::const_iterator) const
386 //std::cout << "adding annot: " << begin << "|" << end << "|" << name << "|" << type << std::endl;
387 annot_list.push_back(annot(begin, end, name, type));
392 struct push_back_seq {
393 std::list<Sequence>& seq_list;
398 push_back_seq(std::list<Sequence>& seq_list_,
402 : seq_list(seq_list_),
409 void operator()(std::string::const_iterator,
410 std::string::const_iterator) const
412 // filter out newlines from our sequence
414 for(std::string::const_iterator seq_i = seq.begin();
418 if (*seq_i != '\015' && *seq_i != '\012') new_seq += *seq_i;
420 //std::cout << "adding seq: " << name << " " << new_seq << std::endl;
423 s.set_fasta_header(name);
424 seq_list.push_back(s);
430 Sequence::parse_annot(std::string data, int start_index, int end_index)
437 std::list<annot> parsed_annots;
438 std::list<Sequence> query_seqs;
441 bool ok = spirit::parse(data.begin(), data.end(),
448 )[spirit::assign_a(species)] >>
452 ( // ignore html tags
453 *(spirit::space_p) >>
455 +(~spirit::ch_p('>')) >>
460 ( // parse an absolute location name
461 (spirit::uint_p[spirit::assign_a(start)] >>
463 spirit::uint_p[spirit::assign_a(end)] >>
468 )[spirit::assign_a(name)] >>
475 )[spirit::assign_a(type)]
477 // to understand how this group gets set
478 // read the comment above struct push_back_annot
479 )[push_back_annot(parsed_annots, start, end, type, name, parsed)]
481 ((spirit::ch_p('>')|spirit::str_p(">")) >>
482 (*(spirit::print_p))[spirit::assign_a(name)] >>
484 (+(spirit::chset<>(Alphabet::nucleic_cstr)))[spirit::assign_a(seq)]
485 )[push_back_seq(query_seqs, name, seq, parsed)]
492 std::stringstream msg;
493 msg << "Error parsing annotation #" << parsed;
494 throw annotation_load_error(msg.str());
496 // add newly parsed annotations to our sequence
497 std::copy(parsed_annots.begin(), parsed_annots.end(), std::back_inserter(annots));
498 // go seearch for query sequences
499 find_sequences(query_seqs.begin(), query_seqs.end());
502 void Sequence::add_annotation(const annot& a)
507 const std::list<annot>& Sequence::annotations() const
512 void Sequence::copy_children(Sequence &new_seq, size_type start, size_type count) const
514 new_seq.motif_list = motif_list;
515 new_seq.annots.clear();
517 for(std::list<annot>::const_iterator annot_i = annots.begin();
518 annot_i != annots.end();
521 size_type annot_begin= annot_i->begin;
522 size_type annot_end = annot_i->end;
524 if (annot_begin < start+count) {
525 if (annot_begin >= start) {
526 annot_begin -= start;
531 if (annot_end < start+count) {
537 annot new_annot(annot_begin, annot_end, annot_i->type, annot_i->name);
538 new_seq.annots.push_back(new_annot);
544 Sequence::subseq(size_type start, size_type count, SeqSpan::strand_type strand) const
546 // FIXME: should i really allow a subsequence of an empty sequence?
552 Sequence new_seq = *this;
553 new_seq.seq = seq->subseq(start, count, strand);
555 copy_children(new_seq, start, count);
561 // FIXME: This needs to be moved into SeqSpan
562 Sequence Sequence::rev_comp() const
564 // a reverse complement is the whole opposite strand
565 return subseq(0, npos, SeqSpan::OppositeStrand);
568 const Alphabet& Sequence::get_alphabet() const
570 return (seq) ? seq->get_alphabet() : Alphabet::empty_alphabet();
573 void Sequence::set_fasta_header(std::string header_)
578 void Sequence::set_species(const std::string& name)
583 std::string Sequence::get_species() const
590 Sequence::get_fasta_header() const
596 Sequence::get_name() const
598 if (header.size() > 0)
600 else if (species.size() > 0)
606 void Sequence::set_sequence(const std::string& s, AlphabetRef a)
608 set_filtered_sequence(s, a, 0, s.size(), SeqSpan::PlusStrand);
611 std::string Sequence::get_sequence() const
613 return seq->sequence();
616 Sequence::const_reference Sequence::operator[](Sequence::size_type i) const
632 Sequence::save(fs::fstream &save_file)
635 std::list<annot>::iterator annots_i;
637 // not sure why, or if i'm doing something wrong, but can't seem to pass
638 // file pointers down to this method from the mussa control class
639 // so each call to save a sequence appends to the file started by mussa_class
640 //save_file.open(save_file_path.c_str(), std::ios::app);
642 save_file << "<Sequence>" << std::endl;
643 save_file << *this << std::endl;
644 save_file << "</Sequence>" << std::endl;
646 save_file << "<Annotations>" << std::endl;
647 save_file << species << std::endl;
648 for (annots_i = annots.begin(); annots_i != annots.end(); ++annots_i)
650 save_file << annots_i->begin << " " << annots_i->end << " " ;
651 save_file << annots_i->name << " " << annots_i->type << std::endl;
653 save_file << "</Annotations>" << std::endl;
658 Sequence::load_museq(fs::path load_file_path, int seq_num)
660 fs::fstream load_file;
661 std::string file_data_line;
664 std::string::size_type space_split_i;
665 std::string annot_value;
668 load_file.open(load_file_path, std::ios::in);
671 // search for the seq_num-th sequence
672 while ( (!load_file.eof()) && (seq_counter < seq_num) )
674 getline(load_file,file_data_line);
675 if (file_data_line == "<Sequence>")
678 getline(load_file, file_data_line);
679 // looks like the sequence is written as a single line
680 set_filtered_sequence(file_data_line, reduced_dna_alphabet, 0, file_data_line.size(), SeqSpan::PlusStrand);
681 getline(load_file, file_data_line);
682 getline(load_file, file_data_line);
683 if (file_data_line == "<Annotations>")
685 getline(load_file, file_data_line);
686 species = file_data_line;
687 while ( (!load_file.eof()) && (file_data_line != "</Annotations>") )
689 getline(load_file,file_data_line);
690 if ((file_data_line != "") && (file_data_line != "</Annotations>"))
692 // need to get 4 values...almost same code 4 times...
693 // get annot start index
694 space_split_i = file_data_line.find(" ");
695 annot_value = file_data_line.substr(0,space_split_i);
696 an_annot.begin = atoi (annot_value.c_str());
697 file_data_line = file_data_line.substr(space_split_i+1);
698 // get annot end index
699 space_split_i = file_data_line.find(" ");
700 annot_value = file_data_line.substr(0,space_split_i);
701 an_annot.end = atoi (annot_value.c_str());
703 if (space_split_i == std::string::npos) // no entry for type or name
705 std::cout << "seq, annots - no type or name\n";
709 else // else get annot type
711 file_data_line = file_data_line.substr(space_split_i+1);
712 space_split_i = file_data_line.find(" ");
713 annot_value = file_data_line.substr(0,space_split_i);
714 an_annot.type = annot_value;
715 if (space_split_i == std::string::npos) // no entry for name
717 std::cout << "seq, annots - no name\n";
720 else // get annot name
722 file_data_line = file_data_line.substr(space_split_i+1);
723 space_split_i = file_data_line.find(" ");
724 annot_value = file_data_line.substr(0,space_split_i);
725 an_annot.type = annot_value;
728 annots.push_back(an_annot); // don't forget to actually add the annot
730 //std::cout << "seq, annots: " << an_annot.start << ", " << an_annot.end
731 // << "-->" << an_annot.type << "::" << an_annot.name << std::endl;
738 void Sequence::add_motif(const Sequence& a_motif)
740 std::vector<int> motif_starts = find_motif(a_motif);
742 for(std::vector<int>::iterator motif_start_i = motif_starts.begin();
743 motif_start_i != motif_starts.end();
746 motif_list.push_back(motif(*motif_start_i, a_motif.get_sequence()));
750 void Sequence::clear_motifs()
755 const std::list<motif>& Sequence::motifs() const
761 Sequence::find_motif(const Sequence& a_motif) const
763 std::vector<int> motif_match_starts;
764 Sequence norm_motif_rc;
766 motif_match_starts.clear();
767 // std::cout << "motif is: " << norm_motif << std::endl;
769 if (a_motif.size() > 0)
771 //std::cout << "Sequence: none blank motif\n";
772 motif_scan(a_motif, &motif_match_starts);
774 norm_motif_rc = a_motif.rev_comp();;
775 // make sure not to do search again if it is a palindrome
776 if (norm_motif_rc != a_motif) {
777 motif_scan(norm_motif_rc, &motif_match_starts);
780 return motif_match_starts;
784 Sequence::motif_scan(const Sequence& a_motif, std::vector<int> * motif_match_starts) const
786 // if there's no sequence we can't scan for it?
787 // should this throw an exception?
790 std::string::size_type seq_i = 0;
791 Sequence::size_type motif_i = 0;
792 Sequence::size_type motif_len = a_motif.length();
793 Sequence::value_type motif_char;
794 Sequence::value_type seq_char;
796 while (seq_i < size())
798 // this is pretty much a straight translation of Nora's python code
799 // to match iupac letter codes
800 motif_char = toupper(a_motif[motif_i]);
801 seq_char = toupper(seq->at(seq_i));
802 if (motif_char =='N')
804 else if (motif_char == seq_char)
806 else if ((motif_char =='M') && ((seq_char=='A') || (seq_char=='C')))
808 else if ((motif_char =='R') && ((seq_char=='A') || (seq_char=='G')))
810 else if ((motif_char =='W') && ((seq_char=='A') || (seq_char=='T')))
812 else if ((motif_char =='S') && ((seq_char=='C') || (seq_char=='G')))
814 else if ((motif_char =='Y') && ((seq_char=='C') || (seq_char=='T')))
816 else if ((motif_char =='K') && ((seq_char=='G') || (seq_char=='T')))
818 else if ((motif_char =='V') &&
819 ((seq_char=='A') || (seq_char=='C') || (seq_char=='G')))
821 else if ((motif_char =='H') &&
822 ((seq_char=='A') || (seq_char=='C') || (seq_char=='T')))
824 else if ((motif_char =='D') &&
825 ((seq_char=='A') || (seq_char=='G') || (seq_char=='T')))
827 else if ((motif_char =='B') &&
828 ((seq_char=='C') || (seq_char=='G') || (seq_char=='T')))
832 // if a motif doesn't match, erase our current trial and try again
837 // end Nora stuff, now we see if a match is found this pass
838 if (motif_i == motif_len)
841 motif_match_starts->push_back(seq_i - motif_len + 1);
847 //std::cout << std::endl;
850 void Sequence::add_string_annotation(std::string a_seq,
853 std::vector<int> seq_starts = find_motif(a_seq);
855 //std::cout << "searching for " << a_seq << " found " << seq_starts.size() << std::endl;
857 for(std::vector<int>::iterator seq_start_i = seq_starts.begin();
858 seq_start_i != seq_starts.end();
861 annots.push_back(annot(*seq_start_i,
862 *seq_start_i+a_seq.size(),
868 void Sequence::find_sequences(std::list<Sequence>::iterator start,
869 std::list<Sequence>::iterator end)
871 while (start != end) {
872 add_string_annotation(start->get_sequence(), start->get_fasta_header());
878 std::ostream& operator<<(std::ostream& out, const Sequence& s)
881 for(Sequence::const_iterator s_i = s.begin(); s_i != s.end(); ++s_i) {
888 bool operator<(const Sequence& x, const Sequence& y)
890 Sequence::const_iterator x_i = x.begin();
891 Sequence::const_iterator y_i = y.begin();
892 // for sequences there's some computation associated with computing .end
894 Sequence::const_iterator xend = x.end();
895 Sequence::const_iterator yend = y.end();
897 if( x_i == xend and y_i == yend ) {
899 } else if ( x_i == xend ) {
901 } else if ( y_i == yend ) {
903 } else if ( (*x_i) < (*y_i)) {
905 } else if ( (*x_i) > (*y_i) ) {
914 template <typename Iter1, typename Iter2>
916 bool sequence_insensitive_equality(Iter1 abegin, Iter1 aend, Iter2 bbegin, Iter2 bend)
918 Iter1 aseq_i = abegin;
919 Iter2 bseq_i = bbegin;
920 if (aend-abegin == bend-bbegin) {
921 // since the length of the two sequences is equal, we only need to
923 for(; aseq_i != aend; ++aseq_i, ++bseq_i) {
924 if (toupper(*aseq_i) != toupper(*bseq_i)) {
934 bool operator==(const Sequence& x, const Sequence& y)
936 if (x.seq and y.seq) {
937 // both x and y are defined
938 if (SeqSpan::isFamily(x.seq, y.seq)) {
939 // both are part of the same SeqSpan tree
940 return *(x.seq) == *(y.seq);
942 // we'll have to do a real comparison
943 return sequence_insensitive_equality<SeqSpan::const_iterator, SeqSpan::const_iterator>(
949 // true if they're both empty (with either a null SeqSpanRef or
950 // a zero length string
951 return (x.size() == y.size());
955 bool operator!=(const Sequence& x, const Sequence& y)
957 return not operator==(x, y);