// Simplify the graph by compacting edges in the given direction
void Bigraph::simplify(EdgeDir dir)
{
bool graph_changed = true;
while(graph_changed)
{
graph_changed = false;
VertexPtrMapIter iter = m_vertices.begin();
while(iter != m_vertices.end())
{
// Get the edges for this direction
EdgePtrVec edges = iter->second->getEdges(dir);
// If there is a single edge in this direction, merge the vertices
// Don't merge singular self edges though
if(edges.size() == 1 && !edges.front()->isSelf())
{
// Check that the edge back is singular as well
Edge* pSingle = edges.front();
Edge* pTwin = pSingle->getTwin();
Vertex* pV2 = pSingle->getEnd();
if(pV2->countEdges(pTwin->getDir()) == 1)
{
merge(iter->second, pSingle);
graph_changed = true;
}
}
++iter;
}
}
}
// Construct the walk structure from a vector of edges
SGWalk::SGWalk(const EdgePtrVec& edgeVec, bool bIndexWalk) : m_extensionDistance(0), m_extensionFinished(false)
{
assert(!edgeVec.empty());
if(bIndexWalk)
m_pWalkIndex = new WalkIndex;
else
m_pWalkIndex = NULL;
// The start vector is the start vertex of the first edge
Edge* first = edgeVec.front();
m_pStartVertex = first->getStart();
for(EdgePtrVec::const_iterator iter = edgeVec.begin();
iter != edgeVec.end();
++iter)
{
addEdge(*iter);
}
}
// Find bubbles (nodes where there is a split and then immediate rejoin) and mark them for removal
bool SGBubbleVisitor::visit(StringGraph* /*pGraph*/, Vertex* pVertex)
{
bool bubble_found = false;
for(size_t idx = 0; idx < ED_COUNT; idx++)
{
EdgeDir dir = EDGE_DIRECTIONS[idx];
EdgePtrVec edges = pVertex->getEdges(dir);
if(edges.size() > 1)
{
Vertex* pStart = pVertex;
Vertex* pEnd = NULL;
// Check the vertices
for(size_t i = 0; i < edges.size(); ++i)
{
Edge* pVWEdge = edges[i];
Vertex* pWVert = pVWEdge->getEnd();
// Get the edges from w in the same direction
EdgeDir transDir = !pVWEdge->getTwinDir();
EdgePtrVec wEdges = pWVert->getEdges(transDir);
if(pWVert->getColor() == GC_RED)
return false;
// If the bubble has collapsed, there should only be one edge
if(wEdges.size() == 1)
{
Vertex* pBubbleEnd = wEdges.front()->getEnd();
if(pBubbleEnd->getColor() == GC_RED)
return false;
}
}
// Mark the vertices
for(size_t i = 0; i < edges.size(); ++i)
{
Edge* pVWEdge = edges[i];
Vertex* pWVert = pVWEdge->getEnd();
// Get the edges from w in the same direction
EdgeDir transDir = !pVWEdge->getTwinDir();
EdgePtrVec wEdges = pWVert->getEdges(transDir);
// If the bubble has collapsed, there should only be one edge
if(wEdges.size() == 1)
{
Vertex* pBubbleEnd = wEdges.front()->getEnd();
if(pBubbleEnd->getColor() == GC_BLACK)
{
// The endpoint has been visited, set this vertex as needing removal
// and set the endpoint as unvisited
pWVert->setColor(GC_RED);
bubble_found = true;
pEnd = pBubbleEnd;
}
else
{
pBubbleEnd->setColor(GC_BLACK);
pWVert->setColor(GC_BLUE);
}
}
}
// Unmark vertices
for(size_t i = 0; i < edges.size(); ++i)
{
Edge* pVWEdge = edges[i];
Vertex* pWVert = pVWEdge->getEnd();
// Get the edges from w in the same direction
EdgeDir transDir = !pVWEdge->getTwinDir();
EdgePtrVec wEdges = pWVert->getEdges(transDir);
// If the bubble has collapsed, there should only be one edge
if(wEdges.size() == 1)
{
Vertex* pBubbleEnd = wEdges.front()->getEnd();
pBubbleEnd->setColor(GC_WHITE);
}
if(pWVert->getColor() == GC_BLUE)
pWVert->setColor(GC_WHITE);
}
(void)pStart;
(void)pEnd;
if(bubble_found)
{
/*
SGWalkVector walkVector;
SGSearch::findWalks(pStart, pEnd, dir, 1000, 20, walkVector);
if(walkVector.size() == 2)
{
SGWalk& walk1 = walkVector[0];
SGWalk& walk2 = walkVector[1];
int len1 = walk1.getStartToEndDistance();
int len2 = walk2.getStartToEndDistance();
int diff = len1 - len2;
std::string type = "SNP";
if(diff != 0)
{
type = "INDEL";
}
std::cout << "Bubble " << pStart->getID() << " to " << pEnd->getID() << " is a "
<< type << "(d: " << diff << ")\n";
}
*/
++num_bubbles;
}
}
}
return bubble_found;
}
