Path_n_count_pair IRKE::inchworm_step (KmerCounter& kcounter, char direction, Kmer_Occurence_Pair kmer, Kmer_visitor& visitor,
Kmer_visitor& eliminator, unsigned int inchworm_round, unsigned int depth,
float MIN_CONNECTIVITY_RATIO, unsigned int max_recurse) {
// cout << "inchworm_step" << endl;
if (IRKE_COMMON::MONITOR >= 2) {
cerr << "\rinchworm: " << string(1,direction)
<< " A:" << INCHWORM_ASSEMBLY_COUNTER << " "
<< " rnd:" << inchworm_round << " D:" << depth << " ";
}
// check to see if kmer exists. If not, return empty container
Path_n_count_pair best_path_n_pair;
if ( !kmer.second
|| visitor.exists(kmer.first) // visited
|| eliminator.exists(kmer.first) // eliminated
) {
// base case, already visited or kmer doesn't exist.
//cout << kmer << "already visited or doesn't exist. ending recursion at depth: " << depth << endl;
return(best_path_n_pair);
}
visitor.add(kmer.first);
if (PACMAN && depth > 0) {
// cerr << "pacman eliminated kmer: " << kmer << endl;
eliminator.add(kmer.first);
}
if (depth < max_recurse) {
vector<Kmer_Occurence_Pair> kmer_candidates;
if (direction == 'F') {
// forward search
kmer_candidates = kcounter.get_forward_kmer_candidates(kmer.first);
}
else {
// reverse search
kmer_candidates = kcounter.get_reverse_kmer_candidates(kmer.first);
}
bool tie = true;
unsigned int recurse_cap = max_recurse;
unsigned int best_path_length = 0;
while (tie) {
vector<Path_n_count_pair> paths;
for (unsigned int i = 0; i < kmer_candidates.size(); i++) {
Kmer_Occurence_Pair kmer_candidate = kmer_candidates[i];
if (kmer_candidate.second // ) {
&& !visitor.exists(kmer_candidate.first) // avoid creating already visited kmers since they're unvisited below...
&& exceeds_min_connectivity(kcounter, kmer, kmer_candidate, MIN_CONNECTIVITY_RATIO) ) {
//cout << endl << "\ttrying " << kmer_candidate << endl;
// recursive call here for extension
Path_n_count_pair p = inchworm_step(kcounter, direction, kmer_candidate, visitor, eliminator, inchworm_round, depth+1, MIN_CONNECTIVITY_RATIO, recurse_cap);
paths.push_back(p);
visitor.erase(kmer_candidate.first); // un-visiting
}
} // end for kmer
if (paths.size() > 1) {
sort(paths.begin(), paths.end(), compare);
if (paths[0].second == paths[1].second // same cumulative coverage values for both paths.
&&
// check last kmer to be sure they're different.
// Not interested in breaking ties between identically scoring paths that end up at the same kmer.
paths[0].first[0] != paths[1].first[0]
) {
// got tie, two different paths and two different endpoints:
if (IRKE_COMMON::MONITOR >= 3) {
cerr << "Got tie! " << ", score: " << paths[0].second << ", recurse at: " << recurse_cap << endl;
vector<unsigned int> v;
cerr << reconstruct_path_sequence(kcounter, paths[0].first, v) << endl;
cerr << reconstruct_path_sequence(kcounter, paths[1].first, v) << endl;
}
if (paths[0].first.size() > best_path_length) {
recurse_cap++;
best_path_length = paths[0].first.size();
}
else {
// cerr << "not able to delve further into the graph, though... Stopping here." << endl;
tie = false;
}
}
else if ((paths[0].second == paths[1].second // same cumulative coverage values for both paths.
&&
paths[0].first[0] == paths[1].first[0] ) // same endpoint
) {
if (IRKE_COMMON::MONITOR >= 3) {
cerr << "Tied, but two different paths join to the same kmer. Choosing first path arbitrarily." << endl;
}
tie = false;
best_path_n_pair = paths[0];
}
else {
// no tie.
tie = false;
best_path_n_pair = paths[0];
}
}
else if (paths.size() == 1) {
tie = false;
best_path_n_pair = paths[0];
}
else {
// no extensions possible.
tie = false;
}
} // end while tie
}
// add current kmer to path, as long as not the original seed kmer!
if (depth > 0) {
best_path_n_pair.first.push_back(kmer.first);
best_path_n_pair.second += kmer.second;
}
return(best_path_n_pair);
}
Path_n_count_pair IRKE::inchworm_step(KmerCounter &kcounter,
char direction,
Kmer_Occurence_Pair kmer,
Kmer_visitor &visitor,
Kmer_visitor &eliminator,
unsigned int inchworm_round,
unsigned int depth,
float MIN_CONNECTIVITY_RATIO,
unsigned int max_recurse)
{
// cout << "inchworm_step" << endl;
if (IRKE_COMMON::MONITOR >= 2) {
cerr << "\rinchworm: " << string(1, direction)
<< " A:" << INCHWORM_ASSEMBLY_COUNTER << " "
<< " rnd:" << inchworm_round << " D:" << depth << " ";
}
// check to see if kmer exists. If not, return empty container
Path_n_count_pair best_path_n_pair;
best_path_n_pair.second = 0; // init
if ( // !kmer.second ||
visitor.exists(kmer.first) // visited
|| eliminator.exists(kmer.first) // eliminated
) {
if (IRKE_COMMON::MONITOR >= 3) {
cerr << "base case, already visited or kmer doesn't exist." << endl;
cerr << kmer.first << " already visited or doesn't exist. ending recursion at depth: " << depth << endl;
}
return (best_path_n_pair);
}
visitor.add(kmer.first);
if (PACMAN && depth > 0) {
// cerr << "pacman eliminated kmer: " << kmer << endl;
eliminator.add(kmer.first);
}
if (depth < max_recurse) {
vector<Kmer_Occurence_Pair> kmer_candidates;
if (direction == 'F') {
// forward search
kmer_candidates = kcounter.get_forward_kmer_candidates(kmer.first);
}
else {
// reverse search
kmer_candidates = kcounter.get_reverse_kmer_candidates(kmer.first);
}
if (IRKE_COMMON::MONITOR >= 3) {
cerr << "Got " << kmer_candidates.size() << " kmer extension candidates." << endl;
}
bool tie = true;
unsigned int recurse_cap = max_recurse;
unsigned int best_path_length = 0;
while (tie) {
// keep trying to break ties if ties encountered.
// this is done by increasing the allowed recursion depth until the tie is broken.
// Recursion depth set via: recurse_cap and incremented if tie is found
vector<Path_n_count_pair> paths; // to collect all the paths rooting from this point
for (unsigned int i = 0; i < kmer_candidates.size(); i++) {
Kmer_Occurence_Pair kmer_candidate = kmer_candidates[i];
if (kmer_candidate.second &&
!visitor.exists(kmer_candidate
.first) // avoid creating already visited kmers since they're unvisited below...
&& exceeds_min_connectivity(kcounter, kmer, kmer_candidate, MIN_CONNECTIVITY_RATIO)) {
//cout << endl << "\ttrying " << kmer_candidate << endl;
// recursive call here for extension
Path_n_count_pair p = inchworm_step(kcounter,
direction,
kmer_candidate,
visitor,
eliminator,
inchworm_round,
depth + 1,
MIN_CONNECTIVITY_RATIO,
recurse_cap);
if (p.first.size() >= 1) {
// only retain paths that include visited nodes.
paths.push_back(p);
}
visitor.erase(kmer_candidate.first); // un-visiting
}
} // end for kmer
if (paths.size() > 1) {
sort(paths.begin(), paths.end(), compare);
if (IRKE_COMMON::__DEVEL_no_greedy_extend) {
// pick a path at random
int rand_index = rand() % paths.size();
tie = false;
if (IRKE_COMMON::MONITOR) {
cerr << "IRKE_COMMON::__DEVEL_no_greedy_extend -- picking random path index: " << rand_index
<< " from size(): " << paths.size() << endl;
}
best_path_n_pair = paths[rand_index];
}
else if (paths[0].second == paths[1].second // same cumulative coverage values for both paths.
&&
// check last kmer to be sure they're different.
// Not interested in breaking ties between identically scoring paths that end up at the same kmer.
paths[0].first[0] != paths[1].first[0]
) {
// got tie, two different paths and two different endpoints:
if (IRKE_COMMON::MONITOR >= 3) {
cerr << "Got tie! " << ", score: " << paths[0].second << ", recurse at: " << recurse_cap
<< endl;
vector<unsigned int> v;
cerr << reconstruct_path_sequence(kcounter, paths[0].first, v) << endl;
cerr << reconstruct_path_sequence(kcounter, paths[1].first, v) << endl;
}
if (IRKE_COMMON::__DEVEL_no_tie_breaking || recurse_cap >= MAX_RECURSION_HARD_STOP) {
tie = false;
int rand_index = rand() % 2;
if (IRKE_COMMON::MONITOR >= 2) {
cerr << "IRKE_COMMON::__DEVEL_no_tie_breaking, so picking path: " << rand_index
<< " at random." << endl;
}
best_path_n_pair = paths[rand_index];
}
else if (paths[0].first.size() > best_path_length) {
// still making progress in extending to try to break the tie. Keep going.
// note, this is the only test that keeps us in this while loop. (tie stays true)
recurse_cap++;
best_path_length = paths[0].first.size();
}
else {
// cerr << "not able to delve further into the graph, though... Stopping here." << endl;
tie = false;
best_path_n_pair = paths[0]; // pick one
}
}
else if ((paths[0].second == paths[1].second // same cumulative coverage values for both paths.
&&
paths[0].first[0] == paths[1].first[0]) // same endpoint
) {
if (IRKE_COMMON::MONITOR >= 3) {
cerr << "Tied, but two different paths join to the same kmer. Choosing first path arbitrarily."
<< endl;
}
tie = false;
best_path_n_pair = paths[0];
}
else {
// no tie.
tie = false;
best_path_n_pair = paths[0];
}
}
else if (paths.size() == 1) {
tie = false;
best_path_n_pair = paths[0];
}
else {
// no extensions possible.
tie = false;
}
} // end while tie
}
// add current kmer to path, as long as not the original seed kmer!
if (depth > 0) {
best_path_n_pair.first.push_back(kmer.first);
best_path_n_pair.second += kmer.second;
}
return (best_path_n_pair);
}
Path_n_count_pair IRKE::inchworm (KmerCounter& kcounter, char direction, kmer_int_type_t kmer, Kmer_visitor& visitor, float min_connectivity) {
// cout << "inchworm" << endl;
Path_n_count_pair entire_path;
unsigned int inchworm_round = 0;
unsigned long num_total_kmers = kcounter.size();
Kmer_visitor eliminator(kcounter.get_kmer_length(), DOUBLE_STRANDED_MODE);
while (true) {
inchworm_round++;
eliminator.clear();
if (inchworm_round > num_total_kmers) {
throw(string ("Error, inchworm rounds have exceeded the number of possible seed kmers"));
}
if (IRKE_COMMON::MONITOR >= 3) {
cerr << endl << "Inchworm round(" << string(1,direction) << "): " << inchworm_round << " searching kmer: " << kmer << endl;
string kmer_str = kcounter.get_kmer_string(kmer);
cerr << kcounter.describe_kmer(kmer_str) << endl;
}
visitor.erase(kmer); // seed kmer must be not visited already.
Kmer_Occurence_Pair kmer_pair(kmer, kcounter.get_kmer_count(kmer));
Path_n_count_pair best_path = inchworm_step(kcounter, direction, kmer_pair, visitor, eliminator, inchworm_round, 0, min_connectivity, MAX_RECURSION);
if (best_path.second > 0) {
// append info to entire path in reverse order, so starts just after seed kmer
vector<kmer_int_type_t>& kmer_list = best_path.first;
unsigned int num_kmers = kmer_list.size();
int first_index = num_kmers - 1;
int last_index = 0;
if (CRAWL) {
last_index = first_index - CRAWL_LENGTH + 1;
if (last_index < 0) {
last_index = 0;
}
}
for (int i = first_index; i >= last_index; i--) {
kmer_int_type_t kmer_extend = kmer_list[i];
entire_path.first.push_back(kmer_extend);
visitor.add(kmer_extend);
entire_path.second += kcounter.get_kmer_count(kmer_extend);
}
kmer = entire_path.first[ entire_path.first.size() -1 ];
}
else {
// no extension possible
break;
}
}
if (IRKE_COMMON::MONITOR >= 3)
cerr << endl;
return(entire_path);
}
Path_n_count_pair IRKE::inchworm(KmerCounter &kcounter,
char direction,
kmer_int_type_t kmer,
Kmer_visitor &visitor,
float min_connectivity)
{
// cout << "inchworm" << endl;
Path_n_count_pair entire_path;
entire_path.second = 0; // init cumulative path coverage
unsigned int inchworm_round = 0;
unsigned long num_total_kmers = kcounter.size();
Kmer_visitor eliminator(kcounter.get_kmer_length(), DOUBLE_STRANDED_MODE);
while (true) {
if (IRKE_COMMON::__DEVEL_rand_fracture) {
// terminate extension with probability of __DEVEL_rand_fracture_prob
float prob_to_fracture = rand() / (float) RAND_MAX;
//cerr << "prob: " << prob_to_fracture << endl;
if (prob_to_fracture <= IRKE_COMMON::__DEVEL_rand_fracture_prob) {
// cerr << "Fracturing at iworm round: " << inchworm_round << " given P: " << prob_to_fracture << endl;
return (entire_path);
}
}
inchworm_round++;
eliminator.clear();
if (inchworm_round > num_total_kmers) {
throw (string("Error, inchworm rounds have exceeded the number of possible seed kmers"));
}
if (IRKE_COMMON::MONITOR >= 3) {
cerr << endl << "Inchworm round(" << string(1, direction) << "): " << inchworm_round << " searching kmer: "
<< kmer << endl;
string kmer_str = kcounter.get_kmer_string(kmer);
cerr << kcounter.describe_kmer(kmer_str) << endl;
}
visitor.erase(kmer); // seed kmer must be not visited already.
Kmer_Occurence_Pair kmer_pair(kmer, kcounter.get_kmer_count(kmer));
Path_n_count_pair best_path = inchworm_step(kcounter,
direction,
kmer_pair,
visitor,
eliminator,
inchworm_round,
0,
min_connectivity,
MAX_RECURSION);
vector<kmer_int_type_t> &kmer_list = best_path.first;
unsigned int num_kmers = kmer_list.size();
if ((IRKE_COMMON::__DEVEL_zero_kmer_on_use && num_kmers >= 1) || best_path.second > 0) {
// append info to entire path in reverse order, so starts just after seed kmer
int first_index = num_kmers - 1;
int last_index = 0;
if (CRAWL) {
last_index = first_index - CRAWL_LENGTH + 1;
if (last_index < 0) {
last_index = 0;
}
}
for (int i = first_index; i >= last_index; i--) {
kmer_int_type_t kmer_extend = kmer_list[i];
entire_path.first.push_back(kmer_extend);
visitor.add(kmer_extend);
//entire_path.second += kcounter.get_kmer_count(kmer_extend);
// selected here, zero out:
if (IRKE_COMMON::__DEVEL_zero_kmer_on_use) {
kcounter.clear_kmer(kmer_extend);
}
}
kmer = entire_path.first[entire_path.first.size() - 1];
entire_path.second += best_path.second;
}
else {
// no extension possible
break;
}
}
if (IRKE_COMMON::MONITOR >= 3)
cerr << "No extension possible." << endl << endl;
return (entire_path);
}