// Extracts junctions from all the SAM hits (based on REF_SKIPs) in the hit file
// resets the stream when finished.
void get_junctions_from_hits(HitStream& hit_stream,
ReadTable& it,
JunctionSet& junctions)
{
HitsForRead curr_hit_group;
hit_stream.next_read_hits(curr_hit_group);
uint32_t curr_obs_order = it.observation_order(curr_hit_group.insert_id);
while(curr_obs_order != 0xFFFFFFFF)
{
for (size_t i = 0; i < curr_hit_group.hits.size(); ++i)
{
BowtieHit& bh = curr_hit_group.hits[i];
if (!bh.contiguous())
{
junctions_from_alignment(bh, junctions);
}
hit_stream.next_read_hits(curr_hit_group);
curr_obs_order = it.observation_order(curr_hit_group.insert_id);
}
}
hit_stream.reset();
}
void closure_driver(vector<FZPipe>& map1,
vector<FZPipe>& map2,
ifstream& ref_stream,
FILE* juncs_file,
FILE* fusions_out)
{
typedef RefSequenceTable::Sequence Reference;
ReadTable it;
RefSequenceTable rt(true);
BowtieHitFactory hit_factory(it, rt);
std::set<Fusion> fusions;
fprintf (stderr, "Finding near-covered motifs...");
CoverageMapVisitor cov_map_visitor(ref_stream, rt);
uint32_t coverage_attempts = 0;
assert(map1.size() == map2.size());
for (size_t num = 0; num < map1.size(); ++num)
{
HitStream left_hs(map1[num].file, &hit_factory, false, true, false);
HitStream right_hs(map2[num].file, &hit_factory, false, true, false);
HitsForRead curr_left_hit_group;
HitsForRead curr_right_hit_group;
left_hs.next_read_hits(curr_left_hit_group);
right_hs.next_read_hits(curr_right_hit_group);
uint32_t curr_right_obs_order = it.observation_order(curr_left_hit_group.insert_id);
uint32_t curr_left_obs_order = it.observation_order(curr_right_hit_group.insert_id);
while(curr_left_obs_order != VMAXINT32 &&
curr_right_obs_order != VMAXINT32)
{
while (curr_left_obs_order < curr_right_obs_order&&
curr_left_obs_order != VMAXINT32 && curr_right_obs_order != VMAXINT32)
{
// Get hit group
left_hs.next_read_hits(curr_left_hit_group);
curr_left_obs_order = it.observation_order(curr_left_hit_group.insert_id);
}
while (curr_left_obs_order > curr_right_obs_order &&
curr_left_obs_order != VMAXINT32 && curr_right_obs_order != VMAXINT32)
{
// Get hit group
right_hs.next_read_hits(curr_right_hit_group);
curr_right_obs_order = it.observation_order(curr_right_hit_group.insert_id);
}
while (curr_left_obs_order == curr_right_obs_order &&
curr_left_obs_order != VMAXINT32 && curr_right_obs_order != VMAXINT32)
{
if (num == 0)
find_fusion_closure(curr_left_hit_group, curr_right_hit_group, fusions);
if (coverage_attempts++ % 10000 == 0)
fprintf (stderr, "Adding covered motifs from pair %d\n", coverage_attempts);
visit_best_pairing(curr_left_hit_group, curr_right_hit_group, cov_map_visitor);
left_hs.next_read_hits(curr_left_hit_group);
curr_left_obs_order = it.observation_order(curr_left_hit_group.insert_id);
right_hs.next_read_hits(curr_right_hit_group);
curr_right_obs_order = it.observation_order(curr_right_hit_group.insert_id);
}
}
}
cov_map_visitor.finalize();
fprintf (stderr, "done\n");
ClosureJunctionSet fwd_splices;
ClosureJunctionSet rev_splices;
JunctionMapVisitor junc_map_visitor(fwd_splices, rev_splices, cov_map_visitor.finders);
fprintf (stderr, "Searching for closures...");
uint32_t closure_attempts = 0;
for (size_t num = 0; num < map1.size(); ++num)
{
map1[num].rewind();
map2[num].rewind();
HitStream left_hs = HitStream(map1[num].file, &hit_factory, false, true, false);
HitStream right_hs = HitStream(map2[num].file, &hit_factory, false, true, false);
HitsForRead curr_left_hit_group;
HitsForRead curr_right_hit_group;
left_hs.next_read_hits(curr_left_hit_group);
right_hs.next_read_hits(curr_right_hit_group);
uint32_t curr_right_obs_order = it.observation_order(curr_left_hit_group.insert_id);
uint32_t curr_left_obs_order = it.observation_order(curr_right_hit_group.insert_id);
while(curr_left_obs_order != VMAXINT32 &&
curr_right_obs_order != VMAXINT32)
{
while (curr_left_obs_order < curr_right_obs_order &&
curr_left_obs_order != VMAXINT32 && curr_right_obs_order != VMAXINT32)
{
// Get hit group
left_hs.next_read_hits(curr_left_hit_group);
curr_left_obs_order = it.observation_order(curr_left_hit_group.insert_id);
}
while (curr_left_obs_order > curr_right_obs_order &&
curr_left_obs_order != VMAXINT32 && curr_right_obs_order != VMAXINT32)
{
// Get hit group
right_hs.next_read_hits(curr_right_hit_group);
curr_right_obs_order = it.observation_order(curr_right_hit_group.insert_id);
}
while (curr_left_obs_order == curr_right_obs_order &&
curr_left_obs_order != VMAXINT32 && curr_right_obs_order != VMAXINT32)
{
if (closure_attempts++ % 10000 == 0)
fprintf (stderr, "Trying to close pair %d\n", closure_attempts);
visit_best_pairing(curr_left_hit_group, curr_right_hit_group, junc_map_visitor);
left_hs.next_read_hits(curr_left_hit_group);
curr_left_obs_order = it.observation_order(curr_left_hit_group.insert_id);
right_hs.next_read_hits(curr_right_hit_group);
curr_right_obs_order = it.observation_order(curr_right_hit_group.insert_id);
}
}
}
for (size_t num = 0; num < map1.size(); ++num)
{
map1[num].close();
map2[num].close();
}
fprintf(stderr, "%lu Forward strand splices\n", fwd_splices.size());
fprintf(stderr, "%lu Reverse strand splices\n", rev_splices.size());
fprintf (stderr, "done\n");
uint32_t num_potential_splices = 0;
fprintf (stderr, "Reporting possible junctions...");
map<uint32_t, pair<JunctionMapVisitor::JunctionTable, JunctionMapVisitor::JunctionTable> >::iterator f_itr;
f_itr = junc_map_visitor._finders.begin();
ClosureJunctionSet::iterator j_itr;
j_itr = fwd_splices.begin();
while (j_itr != fwd_splices.end())
{
fprintf (juncs_file,"%s\t%u\t%u\t%c\n",
rt.get_name(j_itr->refid),
j_itr->left,j_itr->right,'+');
++num_potential_splices;
++j_itr;
}
j_itr = rev_splices.begin();
while (j_itr != rev_splices.end())
{
fprintf (juncs_file,"%s\t%u\t%u\t%c\n",
rt.get_name(j_itr->refid),
j_itr->left,j_itr->right,'-');
++num_potential_splices;
++j_itr;
}
//accept_all_best_hits(best_status_for_inserts);
fprintf(stderr, "done\n");
fprintf(stderr, "Searched for closures between %d pairs\n", searched);
fprintf(stderr, "Successfully closed %d pairs\n", closed);
fprintf(stderr, "Found %d total possible splices\n", num_potential_splices);
// daehwan
#if 0
fprintf (stderr, "Reporting potential fusions...\n");
if(fusions_out){
for(std::set<Fusion>::iterator itr = fusions.begin(); itr != fusions.end(); ++itr){
const char* ref_name1 = rt.get_name(itr->refid1);
const char* ref_name2 = rt.get_name(itr->refid2);
const char* dir = "";
if (itr->dir == FUSION_FR)
dir = "fr";
else if(itr->dir == FUSION_RF)
dir = "rf";
else
dir = "ff";
fprintf(fusions_out,
"%s\t%d\t%s\t%d\t%s\n",
ref_name1,
itr->left,
ref_name2,
itr->right,
dir);
}
fclose(fusions_out);
}else{
fprintf(stderr, "Failed to open fusions file for writing\n");
}
#endif
}
void best_insert_mappings(uint64_t refid,
ReadTable& it,
/*const string& name,*/
HitList& hits1_in_ref,
HitList& hits2_in_ref,
BestInsertAlignmentTable& best_status_for_inserts,
bool prefer_shorter_pairs)
{
long chucked_for_shorter_pair = 0;
std::set<size_t> marked;
HitList::iterator last_good = hits2_in_ref.begin();
for (size_t i = 0; i < hits1_in_ref.size(); ++i)
{
BowtieHit& h1 = hits1_in_ref[i];
pair<HitList::iterator, HitList::iterator> range_pair;
range_pair = equal_range(last_good, hits2_in_ref.end(),
h1, hit_insert_id_lt);
bool found_hit = false;
if (range_pair.first != range_pair.second)
last_good = range_pair.first;
uint32_t obs_order = it.observation_order(h1.insert_id());
for (HitList::iterator f = range_pair.first;
f != range_pair.second;
++f)
{
BowtieHit& h2 = *f;
if (h1.insert_id() == h2.insert_id())
{
// max mate inner distance (genomic)
int min_mate_inner_dist = inner_dist_mean - inner_dist_std_dev;
if (max_mate_inner_dist == -1)
{
max_mate_inner_dist = inner_dist_mean + inner_dist_std_dev;
}
InsertAlignmentGrade s(h1, h2, min_mate_inner_dist, max_mate_inner_dist);
pair<InsertAlignmentGrade, vector<InsertAlignment> >& insert_best
= best_status_for_inserts[obs_order];
InsertAlignmentGrade& current = insert_best.first;
// Is the new status better than the current best one?
if (current < s)
{
insert_best.second.clear();
current = s;
insert_best.second.push_back(InsertAlignment(refid, &h1, &h2));
}
else if (!(s < current))
{
if (prefer_shorter_pairs && current.num_mapped == 2)
{
pair<int, int> dc = pair_distances(*(insert_best.second[0].left_alignment), *(insert_best.second[0].right_alignment));
pair<int, int> ds = pair_distances(h1,h2);
if (ds.second < dc.second)
{
chucked_for_shorter_pair += insert_best.second.size();
insert_best.second.clear();
current = s;
insert_best.second.push_back(InsertAlignment(refid, &h1, &h2));
}
}
else
{
insert_best.second.push_back(InsertAlignment(refid, &h1, &h2));
}
}
marked.insert(f - hits2_in_ref.begin());
found_hit = true;
}
}
if (!found_hit)
{
pair<InsertAlignmentGrade, vector<InsertAlignment> >& insert_best
= best_status_for_inserts[obs_order];
InsertAlignmentGrade& current = insert_best.first;
InsertAlignmentGrade s(h1);
if (current < s)
{
insert_best.second.clear();
current = s;
insert_best.second.push_back(InsertAlignment(refid, &h1, NULL));
}
else if (! (s < current))
{
insert_best.second.push_back(InsertAlignment(refid, &h1, NULL));
}
}
}
for (size_t i = 0; i < hits2_in_ref.size(); ++i)
{
BowtieHit& h2 = hits2_in_ref[i];
uint32_t obs_order = it.observation_order(h2.insert_id());
pair<InsertAlignmentGrade, vector<InsertAlignment> >& insert_best
= best_status_for_inserts[obs_order];
InsertAlignmentGrade& current = insert_best.first;
InsertAlignmentGrade s(h2);
// Did we include h2 as part of a pairing already, or is this first time
// we've seen it? If so, it's a singleton.
if (marked.find(i) == marked.end())
{
if (current < s)
{
insert_best.second.clear();
current = s;
insert_best.second.push_back(InsertAlignment(refid, NULL, &h2));
}
else if (! (s < current))
{
insert_best.second.push_back(InsertAlignment(refid, NULL, &h2));
}
}
}
fprintf(stderr, "Chucked %ld pairs for shorter pairing of same mates\n", chucked_for_shorter_pair);
}
