// Attemps to find the next significant CpG site. Returns true if one was found
// and flase otherwise.
static bool
read_next_significant_cpg(istream &cpg_stream, GenomicRegion &cpg,
double cutoff, bool &skipped_any, bool &sig_raw,
size_t &test_cov, size_t &test_meth,
size_t &rest_cov, size_t &rest_meth) {
GenomicRegion region;
skipped_any = false;
sig_raw = false;
string cpg_encoding;
while (getline(cpg_stream, cpg_encoding)) {
string record, chrom, name, sign;
size_t position;
double raw_pval, adjusted_pval, corrected_pval;
std::istringstream iss(cpg_encoding);
iss.exceptions(std::ios::failbit);
iss >> chrom >> position >> sign >> name >> raw_pval
>> adjusted_pval >> corrected_pval
>> test_cov >> test_meth >> rest_cov >> rest_meth;
if (0 <= corrected_pval && corrected_pval < cutoff) {
cpg.set_chrom(chrom);
cpg.set_start(position);
cpg.set_end(position + 1);
sig_raw = (0 <= raw_pval && raw_pval < cutoff);
return true;
}
skipped_any = true;
}
return false;
}
/**
* \brief TODO
*/
bool
Exon::sameRegion(const GenomicRegion &r) const {
return ((r.get_chrom() == this->region.get_chrom()) &&
(r.get_start() == this->region.get_start()) &&
(r.get_end() == this->region.get_end()) &&
(r.get_strand() == this->region.get_strand()));
}
/**
* \brief Does this exon partially overlap the given genomic region? This
* function doesn't consider an exact region match (i.e. same chrom,
* start, end, and strand) to be a 'partial overlap'.
*/
bool
Exon::partialOverlap(const GenomicRegion &r) const {
if (r.get_chrom() != this->region.get_chrom()) return false;
if (r.get_strand() != this->region.get_strand()) return false;
// by now they must be one the same chromosome and strand... first see if
// they match exactly
if ((r.get_start() == this->region.get_start()) &&
(r.get_end() == this->region.get_end())) return false;
// not an exact match
return (r.overlaps(this->region));
}
static void
convert_coordinates(const unordered_map<size_t, size_t> &cpgs,
GenomicRegion ®ion) {
const unordered_map<size_t, size_t>::const_iterator
start_itr(cpgs.find(region.get_start()));
const unordered_map<size_t, size_t>::const_iterator
end_itr(cpgs.find(region.get_end()));
if (start_itr == cpgs.end() || end_itr == cpgs.end())
throw runtime_error("could not convert:\n" + region.tostring());
region.set_start(start_itr->second);
region.set_end(end_itr->second);
}
static void
get_chrom(const MappedRead &mr,
const chrom_file_map &chrom_files,
GenomicRegion &chrom_region, string &chrom) {
const chrom_file_map::const_iterator fn(chrom_files.find(mr.r.get_chrom()));
if (fn == chrom_files.end())
throw SMITHLABException("could not find chrom: " + mr.r.get_chrom());
chrom.clear();
read_fasta_file(fn->second, mr.r.get_chrom(), chrom);
if (chrom.empty())
throw SMITHLABException("could not find chrom: " + mr.r.get_chrom());
chrom_region.set_chrom(mr.r.get_chrom());
}
static void
get_chrom(const MappedRead &mr,
const vector<string> &all_chroms,
const unordered_map<string, size_t> &chrom_lookup,
GenomicRegion &chrom_region, string &chrom) {
const unordered_map<string, size_t>::const_iterator
the_chrom(chrom_lookup.find(mr.r.get_chrom()));
if (the_chrom == chrom_lookup.end())
throw runtime_error("could not find chrom: " + mr.r.get_chrom());
chrom = all_chroms[the_chrom->second];
if (chrom.empty())
throw runtime_error("could not find chrom: " + mr.r.get_chrom());
chrom_region.set_chrom(mr.r.get_chrom());
}
static void
get_chrom(const bool VERBOSE, const GenomicRegion &r,
const unordered_map<string, string>& chrom_files,
GenomicRegion &chrom_region, string &chrom) {
const unordered_map<string, string>::const_iterator
fn(chrom_files.find(r.get_chrom()));
if (fn == chrom_files.end())
throw runtime_error("could not find chrom: " + r.get_chrom());
chrom.clear();
read_fasta_file(fn->second, r.get_chrom(), chrom);
if (chrom.empty())
throw runtime_error("could not find chrom: " + r.get_chrom());
else {
chrom_region.set_chrom(r.get_chrom());
}
}
void
merge(istream &cpg_stream, ostream &dmr_stream, double cutoff) {
bool skipped_last_cpg, sig_raw;
GenomicRegion dmr;
dmr.set_name("dmr");
size_t dmr_test_cov = 0; size_t dmr_test_meth = 0;
size_t dmr_rest_cov = 0; size_t dmr_rest_meth = 0;
size_t test_cov = 0; size_t test_meth = 0;
size_t rest_cov = 0; size_t rest_meth = 0;
// Find the first significant CpG, or terminate the function if none exist.
if (!read_next_significant_cpg(cpg_stream, dmr, cutoff, skipped_last_cpg,
sig_raw, test_cov, test_meth, rest_cov, rest_meth))
return;
dmr.set_score(sig_raw);
dmr_test_cov += test_cov;
dmr_test_meth += test_meth;
dmr_rest_cov += rest_cov;
dmr_rest_meth += rest_meth;
GenomicRegion cpg;
cpg.set_name("dmr");
while(read_next_significant_cpg(cpg_stream, cpg, cutoff, skipped_last_cpg,
sig_raw, test_cov, test_meth, rest_cov, rest_meth)) {
if (skipped_last_cpg || cpg.get_chrom() != dmr.get_chrom()) {
if (dmr.get_score() != 0)
dmr_stream << dmr.get_chrom() << '\t'
<< dmr.get_start() << '\t'
<< dmr.get_end() << '\t'
<< dmr.get_name() << '\t'
<< dmr.get_score() << '\t'
<< double(dmr_test_meth)/dmr_test_cov -
double(dmr_rest_meth)/dmr_rest_cov << std::endl;
dmr = cpg;
dmr.set_score(sig_raw);
dmr_test_cov = test_cov;
dmr_test_meth = test_meth;
dmr_rest_cov = rest_cov;
dmr_rest_meth = rest_meth;
} else {
dmr.set_end(cpg.get_end());
dmr.set_score(dmr.get_score() + sig_raw);
dmr_test_cov += test_cov;
dmr_test_meth += test_meth;
dmr_rest_cov += rest_cov;
dmr_rest_meth += rest_meth;
}
}
dmr_stream << dmr << std::endl;
}
void GffSplitLocator::process(std::string& sLine, void* pData) {
if (m_iLines++ == 0)
return;
m_pLineElements->clear();
this->split(sLine, '\t', m_pLineElements);
// chrom = 0, strand = 1, start = 2, end = 3
uint32_t iStart = std::stoi( m_pLineElements->at(1) );
uint32_t iEnd = std::stoi( m_pLineElements->at(2) );
GffEntry* pChrom = m_pGffLoader->getChromosome( &(m_pLineElements->at(0)) );
GenomicRegion* pSplit = new GenomicRegion(iStart, iEnd);
std::vector<GffEntry*>* pGenes = pChrom->findChildrenAt(pSplit, m_pSearchLevel, false);
sSplitResults* pResult = NULL;
std::stringstream oOutLine;
if (pGenes->size() == 1)
{
/**
* A single gene explains this split
*/
pResult = this->queryGeneSame(pGenes->at(0), pSplit);
}
if (pGenes->size() > 1)
{
for (uint32_t i = 0; i < pGenes->size(); ++i)
{
GffEntry* pGene = pGenes->at(i);
sSplitResults* pGeneResult = this->queryGeneSame(pGene, pSplit);
if (pResult == NULL) {
pResult = pGeneResult;
} else {
if ((pResult->bSameGene == false) && (pGeneResult->bSameGene == true))
{
pResult = pGeneResult;
continue;
}
if ((pResult->bExonsOfGene == false) && (pGeneResult->bExonsOfGene == true))
{
pResult = pGeneResult;
continue;
}
if ((pResult->bSameTranscriptInOneGene == false) && (pGeneResult->bSameTranscriptInOneGene == true))
{
pResult = pGeneResult;
continue;
}
}
}
}
if (pResult != NULL)
{
oOutLine << sLine.substr(0, sLine.length()-1) << pResult->toString('\t') << std::endl;
std::string sFullLine = oOutLine.str();
m_pFileWriter->writeDirect( sFullLine );
return;
}
pGenes = pChrom->findChildrenAt(pSplit, m_pSearchLevel, true);
std::vector<GffEntry*>* pRegions = NULL;
GffEntry* pRegion = NULL;
if (pGenes->size() == 1)
{
/**
* A single gene explains this split
*/
GffEntry* pGene = pGenes->at(0);
if (pGene->contains( pSplit->getStart() ))
pRegions = pChrom->findChildrenAt(NULL, pSplit->getEnd());
else
pRegions = pChrom->findChildrenAt(NULL, pSplit->getStart());
if (pRegions->size() > 0)
pRegion = pRegions->at(0);
pResult = this->queryGeneDifferent(pGene, pRegion, pSplit);
}
if (pGenes->size() > 1)
{
for (uint32_t i = 0; i < pGenes->size(); ++i)
{
GffEntry* pGene = pGenes->at(i);
if (pGene->contains( pSplit->getStart() ))
pRegions = pChrom->findChildrenAt(NULL, pSplit->getEnd());
else
pRegions = pChrom->findChildrenAt(NULL, pSplit->getStart());
if (pRegions->size() > 0)
pRegion = pRegions->at(0);
sSplitResults* pGeneResult = this->queryGeneDifferent(pGene, pRegion, pSplit);
if (pResult == NULL) {
pResult = pGeneResult;
} else {
if ((pResult->bSameGene == false) && (pGeneResult->bSameGene == true))
{
pResult = pGeneResult;
continue;
}
if ((pResult->bExonsOfGene == false) && (pGeneResult->bExonsOfGene == true))
{
pResult = pGeneResult;
continue;
}
if ((pResult->bSameTranscriptInOneGene == false) && (pGeneResult->bSameTranscriptInOneGene == true))
{
pResult = pGeneResult;
continue;
}
}
}
}
if (pResult != NULL)
{
oOutLine << sLine.substr(0, sLine.length()-1) << pResult->toString('\t') << std::endl;
std::string sFullLine = oOutLine.str();
m_pFileWriter->writeDirect( sFullLine );
return;
}
m_pLineElements->clear();
}
static bool
succeeds(const string &chrom, const size_t position,
const GenomicRegion &r) {
return r.get_chrom() < chrom ||
(chrom == r.get_chrom() && r.get_end() <= position);
}
static bool
precedes(const string &chrom, const size_t position,
const GenomicRegion &r) {
return chrom < r.get_chrom() ||
(chrom == r.get_chrom() && position < r.get_start());
}
static inline bool
end_less(const GenomicRegion &a, const GenomicRegion &b) {
return a.get_end() < b.get_end();
}
static inline bool
same_end(const GenomicRegion &a, const GenomicRegion &b) {
return a.get_end() == b.get_end();
}
static inline bool
start_leq(const GenomicRegion &a, const GenomicRegion &b) {
return a.get_start() <= b.get_start();
}
static inline bool
strand_less(const GenomicRegion &a, const GenomicRegion &b) {
return a.get_strand() <= b.get_strand();
}
