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)),
84 motif_list(new MotifList)
92 Sequence::Sequence(const char *seq, AlphabetRef alphabet_, SeqSpan::strand_type strand_)
95 motif_list(new MotifList)
97 set_filtered_sequence(seq, alphabet_, 0, npos, strand_);
100 Sequence::Sequence(const std::string& seq,
101 AlphabetRef alphabet_,
102 SeqSpan::strand_type strand_)
105 motif_list(new MotifList)
107 set_filtered_sequence(seq, alphabet_, 0, seq.size(), strand_);
110 Sequence::Sequence(const Sequence& o)
115 motif_list(o.motif_list)
119 Sequence::Sequence(const Sequence* o)
124 motif_list(o->motif_list)
128 Sequence::Sequence(const SequenceRef o)
129 : seq(new SeqSpan(o->seq)),
133 motif_list(o->motif_list)
137 Sequence::Sequence(const SeqSpanRef& seq_ref)
141 motif_list(new MotifList)
145 Sequence &Sequence::operator=(const Sequence& s)
152 motif_list = s.motif_list;
157 static void multiplatform_getline(std::istream& in, std::string& line)
162 while(in.good() and !(c == '\012' or c == '\015') ) {
166 // if we have cr-lf eat it
168 if (c=='\012' or c == '\015') {
173 void Sequence::load_fasta(fs::path file_path, int seq_num, int start_index, int end_index)
175 load_fasta(file_path, reduced_nucleic_alphabet, seq_num, start_index, end_index);
178 //! load a fasta file into a sequence
179 void Sequence::load_fasta(fs::path file_path, AlphabetRef a,
180 int seq_num, int start_index, int end_index)
182 fs::fstream data_file;
183 data_file.open(file_path, std::ios::in);
185 if (!data_file.good())
187 throw mussa_load_error("Sequence File: "+file_path.string()+" not found");
190 load_fasta(data_file, a, seq_num, start_index, end_index);
191 } catch(sequence_empty_error e) {
192 // there doesn't appear to be any sequence
193 // catch and rethrow to include the filename
194 std::stringstream msg;
195 msg << "The selected sequence in "
196 << file_path.native_file_string()
197 << " appears to be empty";
198 throw sequence_empty_error(msg.str());
199 } catch(sequence_empty_file_error e) {
200 std::stringstream errormsg;
201 errormsg << file_path.native_file_string()
202 << " did not have any fasta sequences" << std::endl;
203 throw sequence_empty_file_error(errormsg.str());
204 } catch(sequence_invalid_load_error e) {
205 std::ostringstream msg;
206 msg << file_path.native_file_string();
207 msg << " " << e.what();
208 throw sequence_invalid_load_error(msg.str());
213 void Sequence::load_fasta(std::istream& file,
214 int seq_num, int start_index, int end_index)
216 load_fasta(file, reduced_nucleic_alphabet, seq_num, start_index, end_index);
220 Sequence::load_fasta(std::istream& data_file, AlphabetRef a,
222 int start_index, int end_index)
224 std::string file_data_line;
225 int header_counter = 0;
226 size_t line_counter = 0;
227 bool read_seq = true;
228 std::string rev_comp;
229 std::string sequence_raw;
230 std::string seq_tmp; // holds sequence during basic filtering
231 const Alphabet &alpha = Alphabet::get_alphabet(a);
234 throw mussa_load_error("fasta sequence number is 1 based (can't be 0)");
237 // search for the header of the fasta sequence we want
238 while ( (!data_file.eof()) && (header_counter < seq_num) )
240 multiplatform_getline(data_file, file_data_line);
242 if (file_data_line.substr(0,1) == ">")
246 if (header_counter > 0) {
247 header = file_data_line.substr(1);
251 while ( !data_file.eof() && read_seq ) {
252 multiplatform_getline(data_file,file_data_line);
254 if (file_data_line.substr(0,1) == ">")
257 for (std::string::const_iterator line_i = file_data_line.begin();
258 line_i != file_data_line.end();
261 if(alpha.exists(*line_i)) {
262 sequence_raw += *line_i;
264 std::ostringstream msg;
265 msg << "Unrecognized characters in fasta sequence at line ";
267 throw sequence_invalid_load_error(msg.str());
273 // Lastly, if subselection of the sequence was specified we keep cut out
274 // and only keep that part
275 // end_index = 0 means no end was specified, so cut to the end
277 end_index = sequence_raw.size();
279 // sequence filtering for upcasing agctn and convert non AGCTN to N
280 if (end_index-start_index <= 0) {
281 std::string msg("The selected sequence appears to be empty");
282 throw sequence_empty_error(msg);
284 set_filtered_sequence(sequence_raw, a, start_index, end_index-start_index, SeqSpan::PlusStrand);
286 std::string errormsg("There were no fasta sequences");
287 throw sequence_empty_file_error(errormsg);
291 void Sequence::set_filtered_sequence(const std::string &in_seq,
292 AlphabetRef alphabet_,
295 SeqSpan::strand_type strand_)
298 count = in_seq.size() - start;
300 new_seq.reserve(count);
302 // finally, the actual conversion loop
303 const Alphabet& alpha_impl = Alphabet::get_alphabet(alphabet_); // go get one of our actual alphabets
304 std::string::const_iterator seq_i = in_seq.begin()+start;
305 for(size_type i = 0; i != count; ++i, ++seq_i)
307 if (alpha_impl.exists(*seq_i)) {
308 new_seq.append(1, toupper(*seq_i));
310 new_seq.append(1, 'N');
313 SeqSpanRef new_seq_ref(new SeqSpan(new_seq, alphabet_, strand_));
318 Sequence::load_annot(fs::path file_path, int start_index, int end_index)
320 if (not fs::exists(file_path)) {
321 throw mussa_load_error("Annotation File " + file_path.string() + " was not found");
323 if (fs::is_directory(file_path)) {
324 throw mussa_load_error(file_path.string() +
325 " is a directory, please provide a file for annotations."
328 fs::fstream data_stream(file_path, std::ios::in);
331 throw mussa_load_error("Error loading annotation file " + file_path.string());
334 // so i should probably be passing the parse function some iterators
335 // but the annotations files are (currently) small, so i think i can
336 // get away with loading the whole file into memory
339 while(data_stream.good()) {
346 parse_annot(data, start_index, end_index);
347 } catch(annotation_load_error e) {
348 std::ostringstream msg;
349 msg << file_path.native_file_string()
352 throw annotation_load_error(msg.str());
356 /* If this works, yikes, this is some brain hurting code.
358 * what's going on is that when pb_annot is instantiated it stores references
359 * to begin, end, name, type, declared in the parse function, then
360 * when operator() is called it grabs values from those references
361 * and uses that to instantiate an annot object and append that to our
364 * This weirdness is because the spirit library requires that actions
365 * conform to a specific prototype operator()(IteratorT, IteratorT)
366 * which doesn't provide any useful opportunity for me to actually
367 * grab the results of our parsing.
369 * so I instantiate this structure in order to have a place to grab
373 struct push_back_annot {
374 std::list<annot>& annot_list;
381 push_back_annot(std::list<annot>& annot_list_,
387 : annot_list(annot_list_),
396 void operator()(std::string::const_iterator,
397 std::string::const_iterator) const
399 //std::cout << "adding annot: " << begin << "|" << end << "|" << name << "|" << type << std::endl;
400 annot_list.push_back(annot(begin, end, name, type));
405 struct push_back_seq {
406 std::list<Sequence>& seq_list;
411 push_back_seq(std::list<Sequence>& seq_list_,
415 : seq_list(seq_list_),
422 void operator()(std::string::const_iterator,
423 std::string::const_iterator) const
425 // filter out newlines from our sequence
427 for(std::string::const_iterator seq_i = seq.begin();
431 if (*seq_i != '\015' && *seq_i != '\012') new_seq += *seq_i;
433 //std::cout << "adding seq: " << name << " " << new_seq << std::endl;
436 s.set_fasta_header(name);
437 seq_list.push_back(s);
443 Sequence::parse_annot(std::string data, int start_index, int end_index)
450 std::list<annot> parsed_annots;
451 std::list<Sequence> query_seqs;
454 bool ok = spirit::parse(data.begin(), data.end(),
461 )[spirit::assign_a(species)] >>
465 ( // ignore html tags
466 *(spirit::space_p) >>
468 +(~spirit::ch_p('>')) >>
473 ( // parse an absolute location name
474 (spirit::uint_p[spirit::assign_a(start)] >>
476 spirit::uint_p[spirit::assign_a(end)] >>
481 )[spirit::assign_a(name)] >>
488 )[spirit::assign_a(type)]
490 // to understand how this group gets set
491 // read the comment above struct push_back_annot
492 )[push_back_annot(parsed_annots, start, end, type, name, parsed)]
494 ((spirit::ch_p('>')|spirit::str_p(">")) >>
495 (*(spirit::print_p))[spirit::assign_a(name)] >>
497 (+(spirit::chset<>(Alphabet::nucleic_cstr)))[spirit::assign_a(seq)]
498 )[push_back_seq(query_seqs, name, seq, parsed)]
505 std::stringstream msg;
506 msg << "Error parsing annotation #" << parsed;
507 throw annotation_load_error(msg.str());
509 // add newly parsed annotations to our sequence
510 std::copy(parsed_annots.begin(), parsed_annots.end(), std::back_inserter(annots));
511 // go seearch for query sequences
512 find_sequences(query_seqs.begin(), query_seqs.end());
515 void Sequence::add_annotation(const annot& a)
520 const std::list<annot>& Sequence::annotations() const
525 void Sequence::copy_children(Sequence &new_seq, size_type start, size_type count) const
527 new_seq.motif_list = motif_list;
528 new_seq.annots.clear();
530 for(std::list<annot>::const_iterator annot_i = annots.begin();
531 annot_i != annots.end();
534 size_type annot_begin= annot_i->begin;
535 size_type annot_end = annot_i->end;
537 if (annot_begin < start+count) {
538 if (annot_begin >= start) {
539 annot_begin -= start;
544 if (annot_end < start+count) {
550 annot new_annot(annot_begin, annot_end, annot_i->type, annot_i->name);
551 new_seq.annots.push_back(new_annot);
557 Sequence::subseq(size_type start, size_type count, SeqSpan::strand_type strand) const
559 // FIXME: should i really allow a subsequence of an empty sequence?
565 Sequence new_seq = *this;
566 new_seq.seq = seq->subseq(start, count, strand);
567 if (seq->annotations()) {
568 AnnotationsRef a(new Annotations(*(seq->annotations())));
569 new_seq.seq->setAnnotations(a);
571 copy_children(new_seq, start, count);
577 // FIXME: This needs to be moved into SeqSpan
578 Sequence Sequence::rev_comp() const
580 // a reverse complement is the whole opposite strand
581 return subseq(0, npos, SeqSpan::OppositeStrand);
584 const Alphabet& Sequence::get_alphabet() const
586 return (seq) ? seq->get_alphabet() : Alphabet::empty_alphabet();
589 void Sequence::set_fasta_header(std::string header_)
594 void Sequence::set_species(const std::string& name)
599 std::string Sequence::get_species() const
606 Sequence::get_fasta_header() const
612 Sequence::get_name() const
614 if (header.size() > 0)
616 else if (species.size() > 0)
622 void Sequence::set_sequence(const std::string& s, AlphabetRef a)
624 set_filtered_sequence(s, a, 0, s.size(), SeqSpan::PlusStrand);
627 std::string Sequence::get_sequence() const
629 return seq->sequence();
632 Sequence::const_reference Sequence::operator[](Sequence::size_type i) const
644 motif_list.reset(new MotifList);
648 Sequence::save(fs::fstream &save_file)
651 std::list<annot>::iterator annots_i;
653 // not sure why, or if i'm doing something wrong, but can't seem to pass
654 // file pointers down to this method from the mussa control class
655 // so each call to save a sequence appends to the file started by mussa_class
656 //save_file.open(save_file_path.c_str(), std::ios::app);
658 save_file << "<Sequence>" << std::endl;
659 save_file << *this << std::endl;
660 save_file << "</Sequence>" << std::endl;
662 save_file << "<Annotations>" << std::endl;
663 save_file << species << std::endl;
664 for (annots_i = annots.begin(); annots_i != annots.end(); ++annots_i)
666 save_file << annots_i->begin << " " << annots_i->end << " " ;
667 save_file << annots_i->name << " " << annots_i->type << std::endl;
669 save_file << "</Annotations>" << std::endl;
674 Sequence::load_museq(fs::path load_file_path, int seq_num)
676 fs::fstream load_file;
677 std::string file_data_line;
680 std::string::size_type space_split_i;
681 std::string annot_value;
684 load_file.open(load_file_path, std::ios::in);
687 // search for the seq_num-th sequence
688 while ( (!load_file.eof()) && (seq_counter < seq_num) )
690 getline(load_file,file_data_line);
691 if (file_data_line == "<Sequence>")
694 getline(load_file, file_data_line);
695 // looks like the sequence is written as a single line
696 set_filtered_sequence(file_data_line, reduced_dna_alphabet, 0, file_data_line.size(), SeqSpan::PlusStrand);
697 getline(load_file, file_data_line);
698 getline(load_file, file_data_line);
699 if (file_data_line == "<Annotations>")
701 getline(load_file, file_data_line);
702 species = file_data_line;
703 while ( (!load_file.eof()) && (file_data_line != "</Annotations>") )
705 getline(load_file,file_data_line);
706 if ((file_data_line != "") && (file_data_line != "</Annotations>"))
708 // need to get 4 values...almost same code 4 times...
709 // get annot start index
710 space_split_i = file_data_line.find(" ");
711 annot_value = file_data_line.substr(0,space_split_i);
712 an_annot.begin = atoi (annot_value.c_str());
713 file_data_line = file_data_line.substr(space_split_i+1);
714 // get annot end index
715 space_split_i = file_data_line.find(" ");
716 annot_value = file_data_line.substr(0,space_split_i);
717 an_annot.end = atoi (annot_value.c_str());
719 if (space_split_i == std::string::npos) // no entry for type or name
721 std::cout << "seq, annots - no type or name\n";
725 else // else get annot type
727 file_data_line = file_data_line.substr(space_split_i+1);
728 space_split_i = file_data_line.find(" ");
729 annot_value = file_data_line.substr(0,space_split_i);
730 an_annot.type = annot_value;
731 if (space_split_i == std::string::npos) // no entry for name
733 std::cout << "seq, annots - no name\n";
736 else // get annot name
738 file_data_line = file_data_line.substr(space_split_i+1);
739 space_split_i = file_data_line.find(" ");
740 annot_value = file_data_line.substr(0,space_split_i);
741 an_annot.type = annot_value;
744 annots.push_back(an_annot); // don't forget to actually add the annot
746 //std::cout << "seq, annots: " << an_annot.start << ", " << an_annot.end
747 // << "-->" << an_annot.type << "::" << an_annot.name << std::endl;
754 void Sequence::add_motif(const Sequence& a_motif)
756 std::vector<int> motif_starts = find_motif(a_motif);
758 for(std::vector<int>::iterator motif_start_i = motif_starts.begin();
759 motif_start_i != motif_starts.end();
762 motif_list->push_back(motif(*motif_start_i, a_motif.get_sequence()));
766 void Sequence::clear_motifs()
771 motif_list.reset(new MotifList);
774 const Sequence::MotifList& Sequence::motifs() const
780 Sequence::find_motif(const Sequence& a_motif) const
782 std::vector<int> motif_match_starts;
783 Sequence norm_motif_rc;
785 motif_match_starts.clear();
786 // std::cout << "motif is: " << norm_motif << std::endl;
788 if (a_motif.size() > 0)
790 //std::cout << "Sequence: none blank motif\n";
791 motif_scan(a_motif, &motif_match_starts);
793 norm_motif_rc = a_motif.rev_comp();;
794 // make sure not to do search again if it is a palindrome
795 if (norm_motif_rc != a_motif) {
796 motif_scan(norm_motif_rc, &motif_match_starts);
799 return motif_match_starts;
803 Sequence::motif_scan(const Sequence& a_motif, std::vector<int> * motif_match_starts) const
805 // if there's no sequence we can't scan for it?
806 // should this throw an exception?
809 std::string::size_type seq_i = 0;
810 Sequence::size_type motif_i = 0;
811 Sequence::size_type motif_len = a_motif.length();
812 Sequence::value_type motif_char;
813 Sequence::value_type seq_char;
815 while (seq_i < size())
817 // this is pretty much a straight translation of Nora's python code
818 // to match iupac letter codes
819 motif_char = toupper(a_motif[motif_i]);
820 seq_char = toupper(seq->at(seq_i));
821 if (motif_char =='N')
823 else if (motif_char == seq_char)
825 else if ((motif_char =='M') && ((seq_char=='A') || (seq_char=='C')))
827 else if ((motif_char =='R') && ((seq_char=='A') || (seq_char=='G')))
829 else if ((motif_char =='W') && ((seq_char=='A') || (seq_char=='T')))
831 else if ((motif_char =='S') && ((seq_char=='C') || (seq_char=='G')))
833 else if ((motif_char =='Y') && ((seq_char=='C') || (seq_char=='T')))
835 else if ((motif_char =='K') && ((seq_char=='G') || (seq_char=='T')))
837 else if ((motif_char =='V') &&
838 ((seq_char=='A') || (seq_char=='C') || (seq_char=='G')))
840 else if ((motif_char =='H') &&
841 ((seq_char=='A') || (seq_char=='C') || (seq_char=='T')))
843 else if ((motif_char =='D') &&
844 ((seq_char=='A') || (seq_char=='G') || (seq_char=='T')))
846 else if ((motif_char =='B') &&
847 ((seq_char=='C') || (seq_char=='G') || (seq_char=='T')))
851 // if a motif doesn't match, erase our current trial and try again
856 // end Nora stuff, now we see if a match is found this pass
857 if (motif_i == motif_len)
860 motif_match_starts->push_back(seq_i - motif_len + 1);
866 //std::cout << std::endl;
869 void Sequence::add_string_annotation(std::string a_seq,
872 std::vector<int> seq_starts = find_motif(a_seq);
874 //std::cout << "searching for " << a_seq << " found " << seq_starts.size() << std::endl;
876 for(std::vector<int>::iterator seq_start_i = seq_starts.begin();
877 seq_start_i != seq_starts.end();
880 annots.push_back(annot(*seq_start_i,
881 *seq_start_i+a_seq.size(),
887 void Sequence::find_sequences(std::list<Sequence>::iterator start,
888 std::list<Sequence>::iterator end)
890 while (start != end) {
891 add_string_annotation(start->get_sequence(), start->get_fasta_header());
897 std::ostream& operator<<(std::ostream& out, const Sequence& s)
900 for(Sequence::const_iterator s_i = s.begin(); s_i != s.end(); ++s_i) {
907 bool operator<(const Sequence& x, const Sequence& y)
909 Sequence::const_iterator x_i = x.begin();
910 Sequence::const_iterator y_i = y.begin();
911 // for sequences there's some computation associated with computing .end
913 Sequence::const_iterator xend = x.end();
914 Sequence::const_iterator yend = y.end();
916 if( x_i == xend and y_i == yend ) {
918 } else if ( x_i == xend ) {
920 } else if ( y_i == yend ) {
922 } else if ( (*x_i) < (*y_i)) {
924 } else if ( (*x_i) > (*y_i) ) {
933 template <typename Iter1, typename Iter2>
935 bool sequence_insensitive_equality(Iter1 abegin, Iter1 aend, Iter2 bbegin, Iter2 bend)
937 Iter1 aseq_i = abegin;
938 Iter2 bseq_i = bbegin;
939 if (aend-abegin == bend-bbegin) {
940 // since the length of the two sequences is equal, we only need to
942 for(; aseq_i != aend; ++aseq_i, ++bseq_i) {
943 if (toupper(*aseq_i) != toupper(*bseq_i)) {
953 bool operator==(const Sequence& x, const Sequence& y)
955 if (x.seq and y.seq) {
956 // both x and y are defined
957 if (SeqSpan::isFamily(x.seq, y.seq)) {
958 // both are part of the same SeqSpan tree
959 return *(x.seq) == *(y.seq);
961 // we'll have to do a real comparison
962 return sequence_insensitive_equality<SeqSpan::const_iterator, SeqSpan::const_iterator>(
968 // true if they're both empty (with either a null SeqSpanRef or
969 // a zero length string
970 return (x.size() == y.size());
974 bool operator!=(const Sequence& x, const Sequence& y)
976 return not operator==(x, y);