/** Return the properties of the specified vertex, unless u is
* ambiguous, in which case return the length of the ambiguous
* sequence.
*/
static inline
ContigProperties get(vertex_bundle_t, const Graph& g, ContigNode u)
{
return u.ambiguous()
? ContigProperties(u.length() + opt::k - 1, 0)
: g[u];
}
/** Return the sequence of the specified contig node. The sequence
* may be ambiguous or reverse complemented.
*/
static Sequence sequence(const Contigs& contigs, const ContigNode& id)
{
if (id.ambiguous()) {
string s(id.ambiguousSequence());
if (s.length() < opt::k)
transform(s.begin(), s.end(), s.begin(), ::tolower);
return string(opt::k - 1, 'N') + s;
} else {
const Sequence& seq = contigs[id.id()].seq;
return id.sense() ? reverseComplement(seq) : seq;
}
}
/** Return the sequence of the specified contig node. The sequence
* may be ambiguous or reverse complemented.
*/
static const Sequence getSequence(ContigNode id)
{
if (id.ambiguous()) {
string s(id.ambiguousSequence());
if (s.length() < opt::k)
transform(s.begin(), s.end(), s.begin(), ::tolower);
return string(opt::k - 1, 'N') + s;
} else {
string seq(g_contigs[id.id()]);
return id.sense() ? reverseComplement(seq) : seq;
}
}
/** Identify paths subsumed by the specified path.
* @param overlaps [out] paths that are found to overlap
* @return the ID of the subsuming path
*/
static ContigID identifySubsumedPaths(const Lengths& lengths,
ContigPathMap::const_iterator path1It,
ContigPathMap& paths,
set<ContigID>& out,
set<ContigID>& overlaps)
{
ostringstream vout;
out.clear();
ContigID id(path1It->first);
const ContigPath& path = path1It->second;
if (gDebugPrint)
vout << get(g_contigNames, ContigNode(id, false))
<< '\t' << path << '\n';
for (ContigPath::const_iterator it = path.begin();
it != path.end(); ++it) {
ContigNode pivot = *it;
if (pivot.ambiguous() || pivot.id() == id)
continue;
ContigPathMap::iterator path2It
= paths.find(pivot.contigIndex());
if (path2It == paths.end())
continue;
ContigPath path2 = path2It->second;
if (pivot.sense())
reverseComplement(path2.begin(), path2.end());
ContigPath consensus = align(lengths, path, path2, pivot);
if (consensus.empty())
continue;
if (equalIgnoreAmbiguos(consensus, path)) {
if (gDebugPrint)
vout << get(g_contigNames, pivot)
<< '\t' << path2 << '\n';
out.insert(path2It->first);
} else if (equalIgnoreAmbiguos(consensus, path2)) {
// This path is larger. Use it as the seed.
return identifySubsumedPaths(lengths, path2It, paths, out,
overlaps);
} else if (isCycle(lengths, consensus)) {
// The consensus path is a cycle.
bool isCyclePath1 = isCycle(lengths, path);
bool isCyclePath2 = isCycle(lengths, path2);
if (!isCyclePath1 && !isCyclePath2) {
// Neither path is a cycle.
if (gDebugPrint)
vout << get(g_contigNames, pivot)
<< '\t' << path2 << '\n'
<< "ignored\t" << consensus << '\n';
overlaps.insert(id);
overlaps.insert(path2It->first);
} else {
// At least one path is a cycle.
if (gDebugPrint)
vout << get(g_contigNames, pivot)
<< '\t' << path2 << '\n'
<< "cycle\t" << consensus << '\n';
if (isCyclePath1 && isCyclePath2)
out.insert(path2It->first);
else if (!isCyclePath1)
overlaps.insert(id);
else if (!isCyclePath2)
overlaps.insert(path2It->first);
}
} else {
if (gDebugPrint)
vout << get(g_contigNames, pivot)
<< '\t' << path2 << '\n'
<< "ignored\t" << consensus << '\n';
overlaps.insert(id);
overlaps.insert(path2It->first);
}
}
cout << vout.str();
return id;
}
/** Return true if the contigs are equal or both are ambiguous. */
static bool equalOrBothAmbiguos(const ContigNode& a,
const ContigNode& b)
{
return a == b || (a.ambiguous() && b.ambiguous());
}
/** Return the length of the specified contig in k-mer. */
static unsigned getLength(const Lengths& lengths,
const ContigNode& u)
{
return u.ambiguous() ? u.length()
: lengths.at(u.id());
}
