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_)
94 motif_list(new MotifList)
96 set_filtered_sequence(seq, alphabet_, 0, npos, strand_);
99 Sequence::Sequence(const std::string& seq,
100 AlphabetRef alphabet_,
101 SeqSpan::strand_type strand_)
104 motif_list(new MotifList)
106 set_filtered_sequence(seq, alphabet_, 0, seq.size(), strand_);
109 Sequence::Sequence(const Sequence& o)
114 motif_list(o.motif_list)
118 Sequence::Sequence(const Sequence* o)
123 motif_list(o->motif_list)
127 Sequence::Sequence(const SequenceRef o)
128 : seq(new SeqSpan(o->seq)),
132 motif_list(o->motif_list)
136 Sequence::Sequence(const SeqSpanRef& seq_ref)
140 motif_list(new MotifList)
144 Sequence &Sequence::operator=(const Sequence& s)
151 motif_list = s.motif_list;
156 static void multiplatform_getline(std::istream& in, std::string& line)
161 while(in.good() and !(c == '\012' or c == '\015') ) {
165 // if we have cr-lf eat it
167 if (c=='\012' or c == '\015') {
172 void Sequence::load_fasta(fs::path file_path, int seq_num, int start_index, int end_index)
174 load_fasta(file_path, reduced_nucleic_alphabet, seq_num, start_index, end_index);
177 //! load a fasta file into a sequence
178 void Sequence::load_fasta(fs::path file_path, AlphabetRef a,
179 int seq_num, int start_index, int end_index)
181 fs::fstream data_file;
182 data_file.open(file_path, std::ios::in);
184 if (!data_file.good())
186 throw mussa_load_error("Sequence File: "+file_path.string()+" not found");
189 load_fasta(data_file, a, seq_num, start_index, end_index);
190 } catch(sequence_empty_error e) {
191 // there doesn't appear to be any sequence
192 // catch and rethrow to include the filename
193 std::stringstream msg;
194 msg << "The selected sequence in "
195 << file_path.native_file_string()
196 << " appears to be empty";
197 throw sequence_empty_error(msg.str());
198 } catch(sequence_empty_file_error e) {
199 std::stringstream errormsg;
200 errormsg << file_path.native_file_string()
201 << " did not have any fasta sequences" << std::endl;
202 throw sequence_empty_file_error(errormsg.str());
203 } catch(sequence_invalid_load_error e) {
204 std::ostringstream msg;
205 msg << file_path.native_file_string();
206 msg << " " << e.what();
207 throw sequence_invalid_load_error(msg.str());
212 void Sequence::load_fasta(std::istream& file,
213 int seq_num, int start_index, int end_index)
215 load_fasta(file, reduced_nucleic_alphabet, seq_num, start_index, end_index);
219 Sequence::load_fasta(std::istream& data_file, AlphabetRef a,
221 int start_index, int end_index)
223 std::string file_data_line;
224 int header_counter = 0;
225 size_t line_counter = 0;
226 bool read_seq = true;
227 std::string rev_comp;
228 std::string sequence_raw;
229 std::string seq_tmp; // holds sequence during basic filtering
230 const Alphabet &alpha = Alphabet::get_alphabet(a);
233 throw mussa_load_error("fasta sequence number is 1 based (can't be 0)");
236 // search for the header of the fasta sequence we want
237 while ( (!data_file.eof()) && (header_counter < seq_num) )
239 multiplatform_getline(data_file, file_data_line);
241 if (file_data_line.substr(0,1) == ">")
245 if (header_counter > 0) {
246 header = file_data_line.substr(1);
250 while ( !data_file.eof() && read_seq ) {
251 multiplatform_getline(data_file,file_data_line);
253 if (file_data_line.substr(0,1) == ">")
256 for (std::string::const_iterator line_i = file_data_line.begin();
257 line_i != file_data_line.end();
260 if(alpha.exists(*line_i)) {
261 sequence_raw += *line_i;
263 std::ostringstream msg;
264 msg << "Unrecognized characters in fasta sequence at line ";
266 throw sequence_invalid_load_error(msg.str());
272 // Lastly, if subselection of the sequence was specified we keep cut out
273 // and only keep that part
274 // end_index = 0 means no end was specified, so cut to the end
276 end_index = sequence_raw.size();
278 // sequence filtering for upcasing agctn and convert non AGCTN to N
279 if (end_index-start_index <= 0) {
280 std::string msg("The selected sequence appears to be empty");
281 throw sequence_empty_error(msg);
283 set_filtered_sequence(sequence_raw, a, start_index, end_index-start_index, SeqSpan::PlusStrand);
285 std::string errormsg("There were no fasta sequences");
286 throw sequence_empty_file_error(errormsg);
290 void Sequence::set_filtered_sequence(const std::string &in_seq,
291 AlphabetRef alphabet_,
294 SeqSpan::strand_type strand_)
297 count = in_seq.size() - start;
299 new_seq.reserve(count);
301 // finally, the actual conversion loop
302 const Alphabet& alpha_impl = Alphabet::get_alphabet(alphabet_); // go get one of our actual alphabets
303 std::string::const_iterator seq_i = in_seq.begin()+start;
304 for(size_type i = 0; i != count; ++i, ++seq_i)
306 if (alpha_impl.exists(*seq_i)) {
307 new_seq.append(1, toupper(*seq_i));
309 new_seq.append(1, 'N');
312 SeqSpanRef new_seq_ref(new SeqSpan(new_seq, alphabet_, strand_));
317 Sequence::load_annot(fs::path file_path, int start_index, int end_index)
319 if (not fs::exists(file_path)) {
320 throw mussa_load_error("Annotation File " + file_path.string() + " was not found");
322 if (fs::is_directory(file_path)) {
323 throw mussa_load_error(file_path.string() +
324 " is a directory, please provide a file for annotations."
327 fs::fstream data_stream(file_path, std::ios::in);
330 throw mussa_load_error("Error loading annotation file " + file_path.string());
333 // so i should probably be passing the parse function some iterators
334 // but the annotations files are (currently) small, so i think i can
335 // get away with loading the whole file into memory
338 while(data_stream.good()) {
345 parse_annot(data, start_index, end_index);
346 } catch(annotation_load_error e) {
347 std::ostringstream msg;
348 msg << file_path.native_file_string()
351 throw annotation_load_error(msg.str());
355 /* If this works, yikes, this is some brain hurting code.
357 * what's going on is that when pb_annot is instantiated it stores references
358 * to begin, end, name, type, declared in the parse function, then
359 * when operator() is called it grabs values from those references
360 * and uses that to instantiate an annot object and append that to our
363 * This weirdness is because the spirit library requires that actions
364 * conform to a specific prototype operator()(IteratorT, IteratorT)
365 * which doesn't provide any useful opportunity for me to actually
366 * grab the results of our parsing.
368 * so I instantiate this structure in order to have a place to grab
372 struct push_back_annot {
373 std::list<annot>& annot_list;
380 push_back_annot(std::list<annot>& annot_list_,
386 : annot_list(annot_list_),
395 void operator()(std::string::const_iterator,
396 std::string::const_iterator) const
398 //std::cout << "adding annot: " << begin << "|" << end << "|" << name << "|" << type << std::endl;
399 annot_list.push_back(annot(begin, end, name, type));
404 struct push_back_seq {
405 std::list<Sequence>& seq_list;
410 push_back_seq(std::list<Sequence>& seq_list_,
414 : seq_list(seq_list_),
421 void operator()(std::string::const_iterator,
422 std::string::const_iterator) const
424 // filter out newlines from our sequence
426 for(std::string::const_iterator seq_i = seq.begin();
430 if (*seq_i != '\015' && *seq_i != '\012') new_seq += *seq_i;
432 //std::cout << "adding seq: " << name << " " << new_seq << std::endl;
435 s.set_fasta_header(name);
436 seq_list.push_back(s);
442 Sequence::parse_annot(std::string data, int start_index, int end_index)
449 std::list<annot> parsed_annots;
450 std::list<Sequence> query_seqs;
453 bool ok = spirit::parse(data.begin(), data.end(),
460 )[spirit::assign_a(species)] >>
464 ( // ignore html tags
465 *(spirit::space_p) >>
467 +(~spirit::ch_p('>')) >>
472 ( // parse an absolute location name
473 (spirit::uint_p[spirit::assign_a(start)] >>
475 spirit::uint_p[spirit::assign_a(end)] >>
480 )[spirit::assign_a(name)] >>
487 )[spirit::assign_a(type)]
489 // to understand how this group gets set
490 // read the comment above struct push_back_annot
491 )[push_back_annot(parsed_annots, start, end, type, name, parsed)]
493 ((spirit::ch_p('>')|spirit::str_p(">")) >>
494 (*(spirit::print_p))[spirit::assign_a(name)] >>
496 (+(spirit::chset<>(Alphabet::nucleic_cstr)))[spirit::assign_a(seq)]
497 )[push_back_seq(query_seqs, name, seq, parsed)]
504 std::stringstream msg;
505 msg << "Error parsing annotation #" << parsed;
506 throw annotation_load_error(msg.str());
508 // add newly parsed annotations to our sequence
509 std::copy(parsed_annots.begin(), parsed_annots.end(), std::back_inserter(annots));
510 // go seearch for query sequences
511 find_sequences(query_seqs.begin(), query_seqs.end());
514 void Sequence::add_annotation(const annot& a)
519 const std::list<annot>& Sequence::annotations() const
524 void Sequence::copy_children(Sequence &new_seq, size_type start, size_type count) const
526 new_seq.motif_list = motif_list;
527 new_seq.annots.clear();
529 for(std::list<annot>::const_iterator annot_i = annots.begin();
530 annot_i != annots.end();
533 size_type annot_begin= annot_i->begin;
534 size_type annot_end = annot_i->end;
536 if (annot_begin < start+count) {
537 if (annot_begin >= start) {
538 annot_begin -= start;
543 if (annot_end < start+count) {
549 annot new_annot(annot_begin, annot_end, annot_i->type, annot_i->name);
550 new_seq.annots.push_back(new_annot);
556 Sequence::subseq(size_type start, size_type count, SeqSpan::strand_type strand) const
558 // FIXME: should i really allow a subsequence of an empty sequence?
564 Sequence new_seq = *this;
565 new_seq.seq = seq->subseq(start, count, strand);
566 if (seq->annotations()) {
567 AnnotationsRef a(new Annotations(*(seq->annotations())));
568 new_seq.seq->setAnnotations(a);
570 copy_children(new_seq, start, count);
576 // FIXME: This needs to be moved into SeqSpan
577 Sequence Sequence::rev_comp() const
579 // a reverse complement is the whole opposite strand
580 return subseq(0, npos, SeqSpan::OppositeStrand);
583 const Alphabet& Sequence::get_alphabet() const
585 return (seq) ? seq->get_alphabet() : Alphabet::empty_alphabet();
588 void Sequence::set_fasta_header(std::string header_)
593 void Sequence::set_species(const std::string& name)
598 std::string Sequence::get_species() const
605 Sequence::get_fasta_header() const
611 Sequence::get_name() const
613 if (header.size() > 0)
615 else if (species.size() > 0)
621 void Sequence::set_sequence(const std::string& s, AlphabetRef a)
623 set_filtered_sequence(s, a, 0, s.size(), SeqSpan::PlusStrand);
626 std::string Sequence::get_sequence() const
628 return seq->sequence();
631 Sequence::const_reference Sequence::operator[](Sequence::size_type i) const
643 motif_list.reset(new MotifList);
647 Sequence::save(fs::fstream &save_file)
650 std::list<annot>::iterator annots_i;
652 // not sure why, or if i'm doing something wrong, but can't seem to pass
653 // file pointers down to this method from the mussa control class
654 // so each call to save a sequence appends to the file started by mussa_class
655 //save_file.open(save_file_path.c_str(), std::ios::app);
657 save_file << "<Sequence>" << std::endl;
658 save_file << *this << std::endl;
659 save_file << "</Sequence>" << std::endl;
661 save_file << "<Annotations>" << std::endl;
662 save_file << species << std::endl;
663 for (annots_i = annots.begin(); annots_i != annots.end(); ++annots_i)
665 save_file << annots_i->begin << " " << annots_i->end << " " ;
666 save_file << annots_i->name << " " << annots_i->type << std::endl;
668 save_file << "</Annotations>" << std::endl;
673 Sequence::load_museq(fs::path load_file_path, int seq_num)
675 fs::fstream load_file;
676 std::string file_data_line;
679 std::string::size_type space_split_i;
680 std::string annot_value;
683 load_file.open(load_file_path, std::ios::in);
686 // search for the seq_num-th sequence
687 while ( (!load_file.eof()) && (seq_counter < seq_num) )
689 getline(load_file,file_data_line);
690 if (file_data_line == "<Sequence>")
693 getline(load_file, file_data_line);
694 // looks like the sequence is written as a single line
695 set_filtered_sequence(file_data_line, reduced_dna_alphabet, 0, file_data_line.size(), SeqSpan::PlusStrand);
696 getline(load_file, file_data_line);
697 getline(load_file, file_data_line);
698 if (file_data_line == "<Annotations>")
700 getline(load_file, file_data_line);
701 species = file_data_line;
702 while ( (!load_file.eof()) && (file_data_line != "</Annotations>") )
704 getline(load_file,file_data_line);
705 if ((file_data_line != "") && (file_data_line != "</Annotations>"))
707 // need to get 4 values...almost same code 4 times...
708 // get annot start index
709 space_split_i = file_data_line.find(" ");
710 annot_value = file_data_line.substr(0,space_split_i);
711 an_annot.begin = atoi (annot_value.c_str());
712 file_data_line = file_data_line.substr(space_split_i+1);
713 // get annot end index
714 space_split_i = file_data_line.find(" ");
715 annot_value = file_data_line.substr(0,space_split_i);
716 an_annot.end = atoi (annot_value.c_str());
718 if (space_split_i == std::string::npos) // no entry for type or name
720 std::cout << "seq, annots - no type or name\n";
724 else // else get annot type
726 file_data_line = file_data_line.substr(space_split_i+1);
727 space_split_i = file_data_line.find(" ");
728 annot_value = file_data_line.substr(0,space_split_i);
729 an_annot.type = annot_value;
730 if (space_split_i == std::string::npos) // no entry for name
732 std::cout << "seq, annots - no name\n";
735 else // get annot name
737 file_data_line = file_data_line.substr(space_split_i+1);
738 space_split_i = file_data_line.find(" ");
739 annot_value = file_data_line.substr(0,space_split_i);
740 an_annot.type = annot_value;
743 annots.push_back(an_annot); // don't forget to actually add the annot
745 //std::cout << "seq, annots: " << an_annot.start << ", " << an_annot.end
746 // << "-->" << an_annot.type << "::" << an_annot.name << std::endl;
753 void Sequence::add_motif(const Sequence& a_motif)
755 std::vector<int> motif_starts = find_motif(a_motif);
757 for(std::vector<int>::iterator motif_start_i = motif_starts.begin();
758 motif_start_i != motif_starts.end();
761 motif_list->push_back(motif(*motif_start_i, a_motif.get_sequence()));
765 void Sequence::clear_motifs()
770 motif_list.reset(new MotifList);
773 const Sequence::MotifList& Sequence::motifs() const
779 Sequence::find_motif(const Sequence& a_motif) const
781 std::vector<int> motif_match_starts;
782 Sequence norm_motif_rc;
784 motif_match_starts.clear();
785 // std::cout << "motif is: " << norm_motif << std::endl;
787 if (a_motif.size() > 0)
789 //std::cout << "Sequence: none blank motif\n";
790 motif_scan(a_motif, &motif_match_starts);
792 norm_motif_rc = a_motif.rev_comp();;
793 // make sure not to do search again if it is a palindrome
794 if (norm_motif_rc != a_motif) {
795 motif_scan(norm_motif_rc, &motif_match_starts);
798 return motif_match_starts;
802 Sequence::motif_scan(const Sequence& a_motif, std::vector<int> * motif_match_starts) const
804 // if there's no sequence we can't scan for it?
805 // should this throw an exception?
808 std::string::size_type seq_i = 0;
809 Sequence::size_type motif_i = 0;
810 Sequence::size_type motif_len = a_motif.length();
811 Sequence::value_type motif_char;
812 Sequence::value_type seq_char;
814 while (seq_i < size())
816 // this is pretty much a straight translation of Nora's python code
817 // to match iupac letter codes
818 motif_char = toupper(a_motif[motif_i]);
819 seq_char = toupper(seq->at(seq_i));
820 if (motif_char =='N')
822 else if (motif_char == seq_char)
824 else if ((motif_char =='M') && ((seq_char=='A') || (seq_char=='C')))
826 else if ((motif_char =='R') && ((seq_char=='A') || (seq_char=='G')))
828 else if ((motif_char =='W') && ((seq_char=='A') || (seq_char=='T')))
830 else if ((motif_char =='S') && ((seq_char=='C') || (seq_char=='G')))
832 else if ((motif_char =='Y') && ((seq_char=='C') || (seq_char=='T')))
834 else if ((motif_char =='K') && ((seq_char=='G') || (seq_char=='T')))
836 else if ((motif_char =='V') &&
837 ((seq_char=='A') || (seq_char=='C') || (seq_char=='G')))
839 else if ((motif_char =='H') &&
840 ((seq_char=='A') || (seq_char=='C') || (seq_char=='T')))
842 else if ((motif_char =='D') &&
843 ((seq_char=='A') || (seq_char=='G') || (seq_char=='T')))
845 else if ((motif_char =='B') &&
846 ((seq_char=='C') || (seq_char=='G') || (seq_char=='T')))
850 // if a motif doesn't match, erase our current trial and try again
855 // end Nora stuff, now we see if a match is found this pass
856 if (motif_i == motif_len)
859 motif_match_starts->push_back(seq_i - motif_len + 1);
865 //std::cout << std::endl;
868 void Sequence::add_string_annotation(std::string a_seq,
871 std::vector<int> seq_starts = find_motif(a_seq);
873 //std::cout << "searching for " << a_seq << " found " << seq_starts.size() << std::endl;
875 for(std::vector<int>::iterator seq_start_i = seq_starts.begin();
876 seq_start_i != seq_starts.end();
879 annots.push_back(annot(*seq_start_i,
880 *seq_start_i+a_seq.size(),
886 void Sequence::find_sequences(std::list<Sequence>::iterator start,
887 std::list<Sequence>::iterator end)
889 while (start != end) {
890 add_string_annotation(start->get_sequence(), start->get_fasta_header());
896 std::ostream& operator<<(std::ostream& out, const Sequence& s)
899 for(Sequence::const_iterator s_i = s.begin(); s_i != s.end(); ++s_i) {
906 bool operator<(const Sequence& x, const Sequence& y)
908 Sequence::const_iterator x_i = x.begin();
909 Sequence::const_iterator y_i = y.begin();
910 // for sequences there's some computation associated with computing .end
912 Sequence::const_iterator xend = x.end();
913 Sequence::const_iterator yend = y.end();
915 if( x_i == xend and y_i == yend ) {
917 } else if ( x_i == xend ) {
919 } else if ( y_i == yend ) {
921 } else if ( (*x_i) < (*y_i)) {
923 } else if ( (*x_i) > (*y_i) ) {
932 template <typename Iter1, typename Iter2>
934 bool sequence_insensitive_equality(Iter1 abegin, Iter1 aend, Iter2 bbegin, Iter2 bend)
936 Iter1 aseq_i = abegin;
937 Iter2 bseq_i = bbegin;
938 if (aend-abegin == bend-bbegin) {
939 // since the length of the two sequences is equal, we only need to
941 for(; aseq_i != aend; ++aseq_i, ++bseq_i) {
942 if (toupper(*aseq_i) != toupper(*bseq_i)) {
952 bool operator==(const Sequence& x, const Sequence& y)
954 if (x.seq and y.seq) {
955 // both x and y are defined
956 if (SeqSpan::isFamily(x.seq, y.seq)) {
957 // both are part of the same SeqSpan tree
958 return *(x.seq) == *(y.seq);
960 // we'll have to do a real comparison
961 return sequence_insensitive_equality<SeqSpan::const_iterator, SeqSpan::const_iterator>(
967 // true if they're both empty (with either a null SeqSpanRef or
968 // a zero length string
969 return (x.size() == y.size());
973 bool operator!=(const Sequence& x, const Sequence& y)
975 return not operator==(x, y);