void junctions_from_alignment(const BowtieHit& spliced_alignment,
JunctionSet& junctions)
{
vector<pair<Junction, JunctionStats> > juncs;
junctions_from_spliced_hit(spliced_alignment, juncs);
for (size_t i = 0; i < juncs.size(); ++i)
{
pair<Junction, JunctionStats>& junc = juncs[i];
JunctionSet::iterator itr = junctions.find(junc.first);
if (itr != junctions.end())
{
JunctionStats& j = itr->second;
j.left_extent = max(j.left_extent, junc.second.left_extent);
j.right_extent = max(j.right_extent, junc.second.right_extent);
j.min_splice_mms = min(j.min_splice_mms, junc.second.min_splice_mms);
j.supporting_hits++;
}
else
{
assert(junc.first.refid != 0xFFFFFFFF);
junctions[junc.first] = junc.second;
}
}
}
void accept_all_junctions(JunctionSet& junctions,
const uint32_t refid)
{
fprintf(stderr, "Accepting all junctions\n");
for (JunctionSet::iterator itr = junctions.begin(); itr != junctions.end(); ++itr)
{
itr->second.accepted = true;
}
}
void filter_junctions(JunctionSet& junctions)
{
for (JunctionSet::iterator i = junctions.begin(); i != junctions.end(); ++i)
{
accept_if_valid(i->first, i->second);
}
knockout_shadow_junctions(junctions);
}
void validate_junctions(const JunctionSet& junctions)
{
uint32_t invalid_juncs = 0;
for (JunctionSet::const_iterator i = junctions.begin();
i != junctions.end();
++i)
{
if (!i->first.valid())
invalid_juncs++;
}
fprintf(stderr, "Found %d invalid junctions\n", invalid_juncs);
}
/**
* Parse the cigar string of a BowtieHit in order to determine the alignment status.
*/
AlignStatus::AlignStatus(const BowtieHit& bh, const JunctionSet& gtf_junctions)
{
const vector<CigarOp>& cigar = bh.cigar();
_aligned = cigar.size() > 0;
_indelFreeAlignment = true;
_unannotatedSpliceFreeAlignment = true;
_edit_dist = bh.edit_dist();
int j = bh.left();
for (size_t c = 0 ; c < cigar.size(); ++c)
{
Junction junc;
switch(cigar[c].opcode)
{
case REF_SKIP:
junc.refid = bh.ref_id();
junc.left = j;
junc.right = junc.left + cigar[c].length;
junc.antisense = bh.antisense_splice();
j += cigar[c].length;
if (gtf_junctions.find(junc) == gtf_junctions.end())
_unannotatedSpliceFreeAlignment = false;
break;
case MATCH:
j += cigar[c].length;
break;
case DEL:
j += cigar[c].length;
_indelFreeAlignment = false;
break;
case INS:
_indelFreeAlignment = false;
break;
default:
break;
}
}
}
void print_junctions(FILE* junctions_out,
const JunctionSet& junctions,
RefSequenceTable& ref_sequences)
{
uint64_t junc_id = 1;
fprintf(junctions_out, "track name=junctions description=\"TopHat junctions\"\n");
for (JunctionSet::const_iterator i = junctions.begin();
i != junctions.end();
++i)
{
const pair<Junction, JunctionStats>& j_itr = *i;
const Junction& j = j_itr.first;
const JunctionStats& s = j_itr.second;
assert(ref_sequences.get_name(j.refid));
//fprintf(stdout,"%d\t%d\t%d\t%c\n", j.refid, j.left, j.right, j.antisense ? '-' : '+');
print_junction(junctions_out,
ref_sequences.get_name(j.refid),
j,
s,
junc_id++);
}
//fprintf(stderr, "Rejected %d / %d alignments, %d / %d spliced\n", rejected, total, rejected_spliced, total_spliced);
}
void driver(const vector<FILE*>& splice_coords_files,
const vector<FILE*>& insertion_coords_files,
const vector<FILE*>& deletion_coords_files,
ifstream& ref_stream)
{
char splice_buf[2048];
RefSequenceTable rt(true);
JunctionSet junctions;
for (size_t i = 0; i < splice_coords_files.size(); ++i)
{
FILE* splice_coords = splice_coords_files[i];
if (!splice_coords)
continue;
while (fgets(splice_buf, 2048, splice_coords))
{
char* nl = strrchr(splice_buf, '\n');
char* buf = splice_buf;
if (nl) *nl = 0;
/**
Fields are:
1) reference name
2) left coord of splice (last char of the left exon)
3) right coord of splice (first char of the right exon)
*/
char* ref_name = get_token((char**)&buf, "\t");
char* scan_left_coord = get_token((char**)&buf, "\t");
char* scan_right_coord = get_token((char**)&buf, "\t");
char* orientation = get_token((char**)&buf, "\t");
if (!scan_left_coord || !scan_right_coord || !orientation)
{
fprintf(stderr,"Error: malformed splice coordinate record\n");
exit(1);
}
uint32_t ref_id = rt.get_id(ref_name, NULL, 0);
uint32_t left_coord = atoi(scan_left_coord);
uint32_t right_coord = atoi(scan_right_coord);
bool antisense = *orientation == '-';
junctions.insert(make_pair<Junction, JunctionStats>(Junction(ref_id, left_coord, right_coord, antisense), JunctionStats()));
}
}
/*
* Read in the deletion coordinates
* and store in a set
*/
std::set<Deletion> deletions;
for(size_t i=0; i < deletion_coords_files.size(); ++i){
FILE* deletion_coords = deletion_coords_files[i];
if(!deletion_coords){
continue;
}
while (fgets(splice_buf, 2048, deletion_coords))
{
char* nl = strrchr(splice_buf, '\n');
char* buf = splice_buf;
if (nl) *nl = 0;
/**
Fields are:
1) reference name
2) left coord of splice (last char of the left exon)
3) right coord of splice (first char of the right exon)
*/
char* ref_name = get_token((char**)&buf, "\t");
char* scan_left_coord = get_token((char**)&buf, "\t");
char* scan_right_coord = get_token((char**)&buf, "\t");
if (!scan_left_coord || !scan_right_coord)
{
fprintf(stderr,"Error: malformed deletion coordinate record\n");
exit(1);
}
/*
* Note that when reading in a deletion, the left co-ord is the position of the
* first deleted based. Since we are co-opting the junction data structure, need
* to fix up this location
*/
uint32_t ref_id = rt.get_id(ref_name, NULL, 0);
uint32_t left_coord = atoi(scan_left_coord);
uint32_t right_coord = atoi(scan_right_coord);
deletions.insert(Deletion(ref_id, left_coord - 1, right_coord, false));
}
}
/*
* Read in the insertion coordinates
* and store in a set
*/
std::set<Insertion> insertions;
for(size_t i=0; i < insertion_coords_files.size(); ++i){
FILE* insertion_coords = insertion_coords_files[i];
if(!insertion_coords){
continue;
}
while(fgets(splice_buf, 2048, insertion_coords)){
char* nl = strrchr(splice_buf, '\n');
char* buf = splice_buf;
if (nl) *nl = 0;
char* ref_name = get_token((char**)&buf, "\t");
char* scan_left_coord = get_token((char**)&buf, "\t");
char* scan_right_coord = get_token((char**)&buf, "\t");
char* scan_sequence = get_token((char**)&buf, "\t");
if (!scan_left_coord || !scan_sequence || !scan_right_coord)
{
fprintf(stderr,"Error: malformed insertion coordinate record\n");
exit(1);
}
seqan::Dna5String sequence = seqan::Dna5String(scan_sequence);
bool containsN = false;
for(size_t index = 0; index < seqan::length(sequence); index += 1){
/*
* Don't allow any ambiguities in the insertion
*/
if(sequence[index] == 'N'){
containsN = true;
break;
}
}
if(containsN){
continue;
}
seqan::CharString charSequence = sequence;
uint32_t ref_id = rt.get_id(ref_name,NULL,0);
uint32_t left_coord = atoi(scan_left_coord);
insertions.insert(Insertion(ref_id, left_coord, seqan::toCString(charSequence)));
}
}
typedef RefSequenceTable::Sequence Reference;
while(ref_stream.good() &&
!ref_stream.eof())
{
Reference ref_str;
string name;
readMeta(ref_stream, name, Fasta());
string::size_type space_pos = name.find_first_of(" \t\r");
if (space_pos != string::npos)
{
name.resize(space_pos);
}
read(ref_stream, ref_str, Fasta());
uint32_t refid = rt.get_id(name, NULL, 0);
Junction dummy_left(refid, 0, 0, true);
Junction dummy_right(refid, VMAXINT32, VMAXINT32, true);
pair<JunctionSet::iterator, JunctionSet::iterator> r;
r.first = junctions.lower_bound(dummy_left);
r.second = junctions.upper_bound(dummy_right);
JunctionSet::iterator itr = r.first;
while(itr != r.second && itr != junctions.end())
{
print_splice(itr->first, read_length, itr->first.antisense ? "GTAG|rev" : "GTAG|fwd", ref_str, name, cout);
++itr;
}
}
ref_stream.clear();
ref_stream.seekg(0, ios::beg);
while(ref_stream.good() &&
!ref_stream.eof())
{
Reference ref_str;
string name;
readMeta(ref_stream, name, Fasta());
string::size_type space_pos = name.find_first_of(" \t\r");
if (space_pos != string::npos)
{
name.resize(space_pos);
}
read(ref_stream, ref_str, Fasta());
uint32_t refid = rt.get_id(name, NULL,0);
Deletion dummy_left(refid, 0, 0, true);
Deletion dummy_right(refid, VMAXINT32, VMAXINT32, true);
pair<std::set<Deletion>::iterator, std::set<Deletion>::iterator> r;
r.first = deletions.lower_bound(dummy_left);
r.second = deletions.upper_bound(dummy_right);
std::set<Deletion>::iterator itr = r.first;
while(itr != r.second && itr != deletions.end())
{
print_splice((Junction)*itr, read_length, itr->antisense ? "del|rev" : "del|fwd", ref_str, name, cout);
++itr;
}
}
ref_stream.clear();
ref_stream.seekg(0, ios::beg);
while(ref_stream.good() &&
!ref_stream.eof())
{
Reference ref_str;
string name;
readMeta(ref_stream, name, Fasta());
string::size_type space_pos = name.find_first_of(" \t\r");
if (space_pos != string::npos)
{
name.resize(space_pos);
}
read(ref_stream, ref_str, Fasta());
uint32_t refid = rt.get_id(name, NULL,0);
Insertion dummy_left(refid, 0, "");
Insertion dummy_right(refid, VMAXINT32, "");
std::set<Insertion>::iterator itr = insertions.lower_bound(dummy_left);
std::set<Insertion>::iterator upper = insertions.upper_bound(dummy_right);
while(itr != upper && itr != insertions.end()){
print_insertion(*itr, read_length, ref_str, name, cout);
++itr;
}
}
}
void driver(const vector<FILE*>& splice_coords_files,
const vector<FILE*>& insertion_coords_files,
const vector<FILE*>& deletion_coords_files,
const vector<FILE*>& fusion_coords_files,
ifstream& ref_stream)
{
char splice_buf[2048];
RefSequenceTable rt(sam_header, true);
get_seqs(ref_stream, rt, true);
JunctionSet junctions;
for (size_t i = 0; i < splice_coords_files.size(); ++i)
{
FILE* splice_coords = splice_coords_files[i];
if (!splice_coords)
continue;
while (fgets(splice_buf, 2048, splice_coords))
{
char* nl = strrchr(splice_buf, '\n');
char* buf = splice_buf;
if (nl) *nl = 0;
/**
Fields are:
1) reference name
2) left coord of splice (last char of the left exon)
3) right coord of splice (first char of the right exon)
*/
char* ref_name = get_token((char**)&buf, "\t");
char* scan_left_coord = get_token((char**)&buf, "\t");
char* scan_right_coord = get_token((char**)&buf, "\t");
char* orientation = get_token((char**)&buf, "\t");
if (!scan_left_coord || !scan_right_coord || !orientation)
{
fprintf(stderr,"Error: malformed splice coordinate record\n");
exit(1);
}
uint32_t ref_id = rt.get_id(ref_name, NULL, 0);
uint32_t left_coord = atoi(scan_left_coord);
uint32_t right_coord = atoi(scan_right_coord);
bool antisense = *orientation == '-';
junctions.insert(make_pair<Junction, JunctionStats>(Junction(ref_id, left_coord, right_coord, antisense), JunctionStats()));
}
}
/*
* Read in the deletion coordinates
* and store in a set
*/
std::set<Deletion> deletions;
for(size_t i=0; i < deletion_coords_files.size(); ++i){
FILE* deletion_coords = deletion_coords_files[i];
if(!deletion_coords){
continue;
}
while (fgets(splice_buf, 2048, deletion_coords))
{
char* nl = strrchr(splice_buf, '\n');
char* buf = splice_buf;
if (nl) *nl = 0;
/**
Fields are:
1) reference name
2) left coord of splice (last char of the left exon)
3) right coord of splice (first char of the right exon)
*/
char* ref_name = get_token((char**)&buf, "\t");
char* scan_left_coord = get_token((char**)&buf, "\t");
char* scan_right_coord = get_token((char**)&buf, "\t");
if (!scan_left_coord || !scan_right_coord)
{
fprintf(stderr,"Error: malformed deletion coordinate record\n");
exit(1);
}
/*
* Note that when reading in a deletion, the left co-ord is the position of the
* first deleted based. Since we are co-opting the junction data structure, need
* to fix up this location
*/
uint32_t ref_id = rt.get_id(ref_name, NULL, 0);
uint32_t left_coord = atoi(scan_left_coord);
uint32_t right_coord = atoi(scan_right_coord);
deletions.insert(Deletion(ref_id, left_coord - 1, right_coord, false));
}
}
/*
* Read in the insertion coordinates
* and store in a set
*/
std::set<Insertion> insertions;
for(size_t i=0; i < insertion_coords_files.size(); ++i){
FILE* insertion_coords = insertion_coords_files[i];
if(!insertion_coords){
continue;
}
while(fgets(splice_buf, 2048, insertion_coords)){
char* nl = strrchr(splice_buf, '\n');
char* buf = splice_buf;
if (nl) *nl = 0;
char* ref_name = get_token((char**)&buf, "\t");
char* scan_left_coord = get_token((char**)&buf, "\t");
char* scan_right_coord = get_token((char**)&buf, "\t");
char* scan_sequence = get_token((char**)&buf, "\t");
if (!scan_left_coord || !scan_sequence || !scan_right_coord)
{
fprintf(stderr,"Error: malformed insertion coordinate record\n");
exit(1);
}
seqan::Dna5String sequence = seqan::Dna5String(scan_sequence);
bool containsN = false;
for(size_t index = 0; index < seqan::length(sequence); index += 1){
/*
* Don't allow any ambiguities in the insertion
*/
if(sequence[index] == 'N'){
containsN = true;
break;
}
}
if(containsN){
continue;
}
seqan::CharString charSequence = sequence;
uint32_t ref_id = rt.get_id(ref_name,NULL,0);
uint32_t left_coord = atoi(scan_left_coord);
insertions.insert(Insertion(ref_id, left_coord, seqan::toCString(charSequence)));
}
}
std::set<Fusion> fusions;
for(size_t i=0; i < fusion_coords_files.size(); ++i){
FILE* fusion_coords = fusion_coords_files[i];
if(!fusion_coords){
continue;
}
while(fgets(splice_buf, 2048, fusion_coords)){
char* nl = strrchr(splice_buf, '\n');
char* buf = splice_buf;
if (nl) *nl = 0;
char* ref_name1 = strsep((char**)&buf, "\t");
char* scan_left_coord = strsep((char**)&buf, "\t");
char* ref_name2 = strsep((char**)&buf, "\t");
char* scan_right_coord = strsep((char**)&buf, "\t");
char* scan_dir = strsep((char**)&buf, "\t");
if (!ref_name1 || !scan_left_coord || !ref_name2 || !scan_right_coord || !scan_dir)
{
fprintf(stderr,"Error: malformed insertion coordinate record\n");
exit(1);
}
uint32_t ref_id1 = rt.get_id(ref_name1, NULL, 0);
uint32_t ref_id2 = rt.get_id(ref_name2, NULL, 0);
uint32_t left_coord = atoi(scan_left_coord);
uint32_t right_coord = atoi(scan_right_coord);
uint32_t dir = FUSION_FF;
if (strcmp(scan_dir, "fr") == 0)
dir = FUSION_FR;
else if(strcmp(scan_dir, "rf") == 0)
dir = FUSION_RF;
else if(strcmp(scan_dir, "rr") == 0)
dir = FUSION_RR;
fusions.insert(Fusion(ref_id1, ref_id2, left_coord, right_coord, dir));
}
}
{
JunctionSet::iterator itr = junctions.begin();
for (; itr != junctions.end(); ++itr)
{
RefSequenceTable::Sequence* ref_str = rt.get_seq(itr->first.refid);
if (ref_str == NULL) continue;
const char* name = rt.get_name(itr->first.refid);
print_splice(itr->first, read_length, itr->first.antisense ? "GTAG|rev" : "GTAG|fwd", *ref_str, name, cout);
}
}
{
std::set<Deletion>::iterator itr = deletions.begin();
for (; itr != deletions.end(); ++itr)
{
RefSequenceTable::Sequence* ref_str = rt.get_seq(itr->refid);
if (ref_str == NULL) continue;
const char* name = rt.get_name(itr->refid);
print_splice((Junction)*itr, read_length, itr->antisense ? "del|rev" : "del|fwd", *ref_str, name, cout);
}
}
{
std::set<Insertion>::iterator itr = insertions.begin();
for (; itr != insertions.end(); ++itr){
RefSequenceTable::Sequence* ref_str = rt.get_seq(itr->refid);
if (ref_str == NULL) continue;
const char* name = rt.get_name(itr->refid);
print_insertion(*itr, read_length, *ref_str, name, cout);
}
}
{
std::set<Fusion>::iterator itr = fusions.begin();
for (; itr != fusions.end(); ++itr){
RefSequenceTable::Sequence* left_ref_str = rt.get_seq(itr->refid1);
RefSequenceTable::Sequence* right_ref_str = rt.get_seq(itr->refid2);
if (left_ref_str == NULL || right_ref_str == NULL) continue;
const char* left_ref_name = rt.get_name(itr->refid1);
const char* right_ref_name = rt.get_name(itr->refid2);
print_fusion(*itr, read_length, *left_ref_str, *right_ref_str, left_ref_name, right_ref_name, cout);
}
}
}
void knockout_shadow_junctions(JunctionSet& junctions)
{
vector<uint32_t> ref_ids;
for (JunctionSet::iterator i = junctions.begin(); i != junctions.end(); ++i)
{
ref_ids.push_back(i->first.refid);
}
sort(ref_ids.begin(), ref_ids.end());
vector<uint32_t>::iterator new_end = unique(ref_ids.begin(), ref_ids.end());
ref_ids.erase(new_end, ref_ids.end());
for(size_t i = 0; i < ref_ids.size(); ++i)
{
uint32_t refid = ref_ids[i];
Junction dummy_left(refid, 0, 0, true);
Junction dummy_right(refid, 0xFFFFFFFF, 0xFFFFFFFF, true);
pair<JunctionSet::iterator, JunctionSet::iterator> r;
r.first = junctions.lower_bound(dummy_left);
r.second = junctions.upper_bound(dummy_right);
JunctionSet::iterator itr = r.first;
while(itr != r.second && itr != junctions.end())
{
if (itr->second.accepted)
{
Junction fuzzy_left = itr->first;
Junction fuzzy_right = itr->first;
fuzzy_left.left -= min_anchor_len;
fuzzy_right.right += min_anchor_len;
fuzzy_left.antisense = !itr->first.antisense;
fuzzy_right.antisense = !itr->first.antisense;
pair<JunctionSet::iterator, JunctionSet::iterator> s;
s.first = junctions.lower_bound(fuzzy_left);
s.second = junctions.upper_bound(fuzzy_right);
JunctionSet::iterator itr2 = s.first;
int junc_support = itr->second.supporting_hits;
while(itr2 != s.second && itr2 != junctions.end())
{
int left_diff = itr->first.left - itr2->first.left;
int right_diff = itr->first.right - itr2->first.right;
if (itr != itr2 &&
itr->first.antisense != itr2->first.antisense &&
(left_diff < min_anchor_len || right_diff < min_anchor_len))
{
if (junc_support < itr2->second.supporting_hits)
itr->second.accepted = false;
}
++itr2;
}
}
++itr;
}
}
}