vector<kmer_int_type_t> IRKE::build_inchworm_contig_from_seed(kmer_int_type_t kmer, KmerCounter &kcounter,
float min_connectivity, unsigned int &total_counts,
bool)
{
unsigned int kmer_count = kcounter.get_kmer_count(kmer);
/* Extend to the right */
unsigned int kmer_length = kcounter.get_kmer_length();
Kmer_visitor visitor(kmer_length, DOUBLE_STRANDED_MODE);
Path_n_count_pair selected_path_n_pair_forward = inchworm(kcounter, 'F', kmer, visitor, min_connectivity);
visitor.clear();
// add selected path to visitor
vector<kmer_int_type_t> &forward_path = selected_path_n_pair_forward.first;
if (IRKE_COMMON::MONITOR >= 2) {
cerr << "Forward path contains: " << forward_path.size() << " kmers. " << endl;
}
for (unsigned int i = 0; i < forward_path.size(); i++) {
kmer_int_type_t kmer = forward_path[i];
visitor.add(kmer);
if (IRKE_COMMON::MONITOR >= 2) {
cerr << "\tForward path kmer: " << kcounter.get_kmer_string(kmer) << endl;
}
}
/* Extend to the left */
visitor.erase(kmer); // reset the seed
Path_n_count_pair selected_path_n_pair_reverse = inchworm(kcounter, 'R', kmer, visitor, min_connectivity);
if (IRKE_COMMON::MONITOR >= 2) {
vector<kmer_int_type_t> &reverse_path = selected_path_n_pair_reverse.first;
cerr << "Reverse path contains: " << reverse_path.size() << " kmers. " << endl;
for (unsigned int i = 0; i < reverse_path.size(); i++) {
cerr << "\tReverse path kmer: " << kcounter.get_kmer_string(reverse_path[i]) << endl;
}
}
total_counts = selected_path_n_pair_forward.second + selected_path_n_pair_reverse.second + kmer_count;
vector<kmer_int_type_t> &reverse_path = selected_path_n_pair_reverse.first;
vector<kmer_int_type_t> joined_path = _join_forward_n_reverse_paths(reverse_path, kmer, forward_path);
return (joined_path);
}
void IRKE::compute_sequence_assemblies(KmerCounter& kcounter, float min_connectivity,
unsigned int MIN_ASSEMBLY_LENGTH, unsigned int MIN_ASSEMBLY_COVERAGE,
bool WRITE_COVERAGE, string COVERAGE_OUTPUT_FILENAME) {
if (! got_sorted_kmers_flag) {
stringstream error;
error << stacktrace() << " Error, must populate_sorted_kmers_list() before computing sequence assemblies" << endl;
throw(error.str());
}
unsigned int kmer_length = kcounter.get_kmer_length();
ofstream coverage_writer;
if (WRITE_COVERAGE) {
coverage_writer.open(COVERAGE_OUTPUT_FILENAME.c_str());
}
vector<Kmer_counter_map_iterator>& kmers = sorted_kmers; //kcounter.get_kmers_sort_descending_counts();
unsigned long init_size = kcounter.size();
// string s = "before.kmers";
// kcounter.dump_kmers_to_file(s);
for (unsigned int i = 0; i < kmers.size(); i++) {
// cerr << "round: " << i << endl;
unsigned long kmer_counter_size = kcounter.size();
if (kmer_counter_size > init_size) {
// string s = "after.kmers";
// kcounter.dump_kmers_to_file(s);
stringstream error;
error << stacktrace() << "Error, Kcounter size has grown from " << init_size
<< " to " << kmer_counter_size << endl;
throw (error.str());
}
kmer_int_type_t kmer = kmers[i]->first;
unsigned int kmer_count = kmers[i]->second;
if (kmer_count == 0) {
continue;
}
if (IRKE_COMMON::MONITOR >= 2) {
cerr << "SEED kmer: " << kcounter.get_kmer_string(kmer) << ", count: " << kmer_count << endl;
}
if (kmer == revcomp_val(kmer, kmer_length)) {
// palindromic kmer, avoid palindromes as seeds
if (IRKE_COMMON::MONITOR >= 2) {
cerr << "SEED kmer: " << kcounter.get_kmer_string(kmer) << " is palidnromic. Skipping. " << endl;
}
continue;
}
if (kmer_count < MIN_SEED_COVERAGE) {
if (IRKE_COMMON::MONITOR >= 2) {
cerr << "-seed has insufficient coverage, skipping" << endl;
}
continue;
}
float entropy = compute_entropy(kmer, kmer_length);
if (entropy < MIN_SEED_ENTROPY) {
if (IRKE_COMMON::MONITOR >= 2) {
cerr << "-skipping seed due to low entropy: " << entropy << endl;
}
continue;
}
/* Extend to the right */
Kmer_visitor visitor(kmer_length, DOUBLE_STRANDED_MODE);
Path_n_count_pair selected_path_n_pair_forward = inchworm(kcounter, 'F', kmer, visitor, min_connectivity);
visitor.clear();
// add selected path to visitor
vector<kmer_int_type_t>& forward_path = selected_path_n_pair_forward.first;
if (IRKE_COMMON::MONITOR >= 2) {
cerr << "Forward path contains: " << forward_path.size() << " kmers. " << endl;
}
for (unsigned int i = 0; i < forward_path.size(); i++) {
kmer_int_type_t kmer = forward_path[i];
visitor.add(kmer);
if (IRKE_COMMON::MONITOR >= 2) {
cerr << "\tForward path kmer: " << kcounter.get_kmer_string(kmer) << endl;
}
}
/* Extend to the left */
visitor.erase(kmer); // reset the seed
Path_n_count_pair selected_path_n_pair_reverse = inchworm(kcounter, 'R', kmer, visitor, min_connectivity);
if (IRKE_COMMON::MONITOR >= 2) {
vector<kmer_int_type_t>& reverse_path = selected_path_n_pair_reverse.first;
cerr << "Reverse path contains: " << reverse_path.size() << " kmers. " << endl;
for (unsigned int i = 0; i < reverse_path.size(); i++) {
cerr << "\tReverse path kmer: " << kcounter.get_kmer_string(reverse_path[i]) << endl;
}
}
unsigned int total_counts = selected_path_n_pair_forward.second + selected_path_n_pair_reverse.second + kcounter.get_kmer_count(kmer);
vector<kmer_int_type_t>& reverse_path = selected_path_n_pair_reverse.first;
vector<kmer_int_type_t> joined_path = _join_forward_n_reverse_paths(reverse_path, kmer, forward_path);
// report sequence reconstructed from path.
vector<unsigned int> assembly_base_coverage;
string sequence = reconstruct_path_sequence(kcounter, joined_path, assembly_base_coverage);
unsigned int avg_cov = static_cast<unsigned int> ( (float)total_counts/(sequence.length()-kcounter.get_kmer_length() +1) + 0.5);
/*
cout << "Inchworm-reconstructed sequence, length: " << sequence.length()
<< ", avgCov: " << avg_cov
<< " " << sequence << endl;
*/
if (sequence.length() >= MIN_ASSEMBLY_LENGTH && avg_cov >= MIN_ASSEMBLY_COVERAGE) {
INCHWORM_ASSEMBLY_COUNTER++;
stringstream headerstream;
headerstream << ">a" << INCHWORM_ASSEMBLY_COUNTER << ";" << avg_cov
<< " K: " << kmer_length
<< " length: " << sequence.length();
string header = headerstream.str();
sequence = add_fasta_seq_line_breaks(sequence, 60);
cout << header << endl << sequence << endl;
if (WRITE_COVERAGE) {
coverage_writer << header << endl;
for (unsigned int i = 0; i < assembly_base_coverage.size(); i++) {
coverage_writer << assembly_base_coverage[i];
if ( (i+1) % 30 == 0) {
coverage_writer << endl;
}
else {
coverage_writer << " ";
}
}
coverage_writer << endl;
}
}
// remove path
for (unsigned int i = 0; i < joined_path.size(); i++) {
kmer_int_type_t kmer = joined_path[i];
/*
if (DEBUG) {
cout << "\tpruning kmer: " << kmer << endl;
}
*/
kcounter.clear_kmer(kmer);
}
/*
if (DEBUG) {
cout << "done pruning kmers." << endl;
}
*/
}
if (IRKE_COMMON::MONITOR) {
cerr << endl;
}
if (WRITE_COVERAGE) {
coverage_writer.close();
}
// drop sorted kmer list as part of cleanup
clear_sorted_kmers_list();
return; // end of runIRKE
}