void Aligner::align_internal(Alignment& alignment, vector<Alignment>* multi_alignments, Graph& g, int64_t pinned_node_id, bool pin_left, int32_t max_alt_alns, bool print_score_matrices) { // check input integrity if (pin_left && !pinned_node_id) { cerr << "error:[Aligner] cannot choose pinned end in non-pinned alignment" << endl; exit(EXIT_FAILURE); } if (multi_alignments && !pinned_node_id) { cerr << "error:[Aligner] multiple traceback is not valid in local alignment, only pinned and global" << endl; exit(EXIT_FAILURE); } if (!(multi_alignments) && max_alt_alns != 1) { cerr << "error:[Aligner] cannot specify maximum number of alignments in single alignment" << endl; exit(EXIT_FAILURE); } // alignment pinning algorithm is based on pinning in bottom right corner, if pinning in top // left we need to reverse all the sequences first and translate the alignment back later // create reversed objects if necessary Graph reversed_graph; string reversed_sequence; if (pin_left) { reversed_sequence.resize(alignment.sequence().length()); reverse_copy(alignment.sequence().begin(), alignment.sequence().end(), reversed_sequence.begin()); reverse_graph(g, reversed_graph); } // choose forward or reversed objects Graph* align_graph; string* align_sequence; if (pin_left) { align_graph = &reversed_graph; align_sequence = &reversed_sequence; } else { align_graph = &g; align_sequence = alignment.mutable_sequence(); } // convert into gssw graph and get the counterpart to pinned node (if pinning) gssw_node* pinned_node = nullptr; gssw_graph* graph = create_gssw_graph(*align_graph, pinned_node_id, &pinned_node); if (pinned_node_id & !pinned_node) { cerr << "error:[Aligner] pinned node for pinned alignment is not in graph" << endl; exit(EXIT_FAILURE); } // perform dynamic programming gssw_graph_fill(graph, (*align_sequence).c_str(), nt_table, score_matrix, gap_open, gap_extension, 15, 2); // traceback either from pinned position or optimal local alignment if (pinned_node) { // trace back pinned alignment gssw_graph_mapping** gms = gssw_graph_trace_back_pinned_multi (graph, pinned_node, max_alt_alns, (*align_sequence).c_str(), (*align_sequence).size(), nt_table, score_matrix, gap_open, gap_extension); if (pin_left) { // translate graph and mappings into original node space unreverse_graph(reversed_graph); for (int32_t i = 0; i < max_alt_alns; i++) { unreverse_graph_mapping(gms[i]); } } // convert optimal alignment and store it in the input Alignment object (in the multi alignment, // this will have been set to the first in the vector) if (gms[0]->score > 0) { // have a mapping, can just convert normally gssw_mapping_to_alignment(graph, gms[0], alignment, print_score_matrices); } else { // gssw will not identify mappings with 0 score, infer location based on pinning Mapping* mapping = alignment.mutable_path()->add_mapping(); mapping->set_rank(1); // locate at the end of the node Position* position = mapping->mutable_position(); position->set_node_id(pinned_node_id); position->set_offset(pin_left ? 0 : pinned_node->len); // soft clip Edit* edit = mapping->add_edit(); edit->set_to_length(alignment.sequence().length()); edit->set_sequence(alignment.sequence()); } if (multi_alignments) { // determine how many non-null alignments were returned int32_t num_non_null = max_alt_alns; for (int32_t i = 1; i < max_alt_alns; i++) { if (gms[i]->score <= 0) { num_non_null = i; break; } } // reserve to avoid illegal access errors that occur when the vector reallocates multi_alignments->reserve(num_non_null); // copy the primary alignment multi_alignments->emplace_back(alignment); // convert the alternate alignments and store them at the back of the vector (this will not // execute if we are doing single alignment) for (int32_t i = 1; i < num_non_null; i++) { gssw_graph_mapping* gm = gms[i]; // make new alignment object multi_alignments->emplace_back(); Alignment& next_alignment = multi_alignments->back(); // copy over sequence information from the primary alignment next_alignment.set_sequence(alignment.sequence()); next_alignment.set_quality(alignment.quality()); // get path of the alternate alignment gssw_mapping_to_alignment(graph, gm, next_alignment, print_score_matrices); } } for (int32_t i = 0; i < max_alt_alns; i++) { gssw_graph_mapping_destroy(gms[i]); } free(gms); } else { // trace back local alignment gssw_graph_mapping* gm = gssw_graph_trace_back (graph, (*align_sequence).c_str(), (*align_sequence).size(), nt_table, score_matrix, gap_open, gap_extension); gssw_mapping_to_alignment(graph, gm, alignment, print_score_matrices); gssw_graph_mapping_destroy(gm); } //gssw_graph_print_score_matrices(graph, sequence.c_str(), sequence.size(), stderr); gssw_graph_destroy(graph); }
void Aligner::gssw_mapping_to_alignment(gssw_graph* graph, gssw_graph_mapping* gm, Alignment& alignment, bool print_score_matrices) { alignment.clear_path(); alignment.set_score(gm->score); alignment.set_query_position(0); Path* path = alignment.mutable_path(); //alignment.set_cigar(graph_cigar(gm)); gssw_graph_cigar* gc = &gm->cigar; gssw_node_cigar* nc = gc->elements; int to_pos = 0; int from_pos = gm->position; //cerr << "gm->position " << gm->position << endl; string& to_seq = *alignment.mutable_sequence(); //cerr << "-------------" << endl; if (print_score_matrices) { gssw_graph_print_score_matrices(graph, to_seq.c_str(), to_seq.size(), stderr); //cerr << alignment.DebugString() << endl; } for (int i = 0; i < gc->length; ++i, ++nc) { if (i > 0) from_pos = 0; // reset for each node after the first // check that the current alignment has a non-zero length gssw_cigar* c = nc->cigar; int l = c->length; if (l == 0) continue; gssw_cigar_element* e = c->elements; Node* from_node = (Node*) nc->node->data; string& from_seq = *from_node->mutable_sequence(); Mapping* mapping = path->add_mapping(); mapping->mutable_position()->set_node_id(nc->node->id); mapping->mutable_position()->set_offset(from_pos); mapping->set_rank(path->mapping_size()); //cerr << from_node->id() << ":" << endl; for (int j=0; j < l; ++j, ++e) { Edit* edit; int32_t length = e->length; //cerr << e->length << e->type << endl; switch (e->type) { case 'M': case 'X': case 'N': { // do the sequences match? // emit a stream of "SNPs" and matches int h = from_pos; int last_start = from_pos; int k = to_pos; for ( ; h < from_pos + length; ++h, ++k) { //cerr << h << ":" << k << " " << from_seq[h] << " " << to_seq[k] << endl; if (from_seq[h] != to_seq[k]) { // emit the last "match" region if (h-last_start > 0) { edit = mapping->add_edit(); edit->set_from_length(h-last_start); edit->set_to_length(h-last_start); } // set up the SNP edit = mapping->add_edit(); edit->set_from_length(1); edit->set_to_length(1); edit->set_sequence(to_seq.substr(k,1)); last_start = h+1; } } // handles the match at the end or the case of no SNP if (h-last_start > 0) { edit = mapping->add_edit(); edit->set_from_length(h-last_start); edit->set_to_length(h-last_start); } to_pos += length; from_pos += length; } break; case 'D': edit = mapping->add_edit(); edit->set_from_length(length); edit->set_to_length(0); from_pos += length; break; case 'I': edit = mapping->add_edit(); edit->set_from_length(0); edit->set_to_length(length); edit->set_sequence(to_seq.substr(to_pos, length)); to_pos += length; break; case 'S': // note that soft clips and insertions are semantically equivalent // and can only be differentiated by their position in the read // with soft clips coming at the start or end edit = mapping->add_edit(); edit->set_from_length(0); edit->set_to_length(length); edit->set_sequence(to_seq.substr(to_pos, length)); to_pos += length; break; default: cerr << "error:[Aligner::gssw_mapping_to_alignment] " << "unsupported cigar op type " << e->type << endl; exit(1); break; } } //cerr << "path to_length " << path_to_length(*path) << endl; } // set identity alignment.set_identity(identity(alignment.path())); }