void PoaGraphImpl::tracebackAndThread(std::string sequence,
const AlignmentColumnMap& alignmentColumnForVertex,
AlignMode alignMode, std::vector<Vertex>* outputPath)
{
const int I = sequence.length();
// perform traceback from (I,$), threading the new sequence into
// the graph as we go.
int i = I;
const AlignmentColumn* curCol;
VD v = null_vertex, forkVertex = null_vertex;
VD u = exitVertex_;
VD startSpanVertex;
VD endSpanVertex = alignmentColumnForVertex.at(exitVertex_)->PreviousVertex[I];
if (outputPath) {
outputPath->resize(I);
std::fill(outputPath->begin(), outputPath->end(), (size_t)-1);
}
#define READPOS (i - 1)
#define VERTEX_ON_PATH(readPos, v) \
if (outputPath) { \
(*outputPath)[(readPos)] = externalize(v); \
}
while (!(u == enterVertex_ && i == 0)) {
// u -> v
// u: current vertex
// v: vertex last visited in traceback (could be == u)
// forkVertex: the vertex that will be the target of a new edge
Vertex uExt = this->externalize(u); // DEBUGGING
Vertex vExt = this->externalize(v); // DEBUGGING
curCol = alignmentColumnForVertex.at(u);
assert(curCol != NULL);
PoaNode& curNodeInfo = vertexInfoMap_[u];
VD prevVertex = curCol->PreviousVertex[i];
MoveType reachingMove = curCol->ReachingMove[i];
if (reachingMove == StartMove) {
assert(v != null_vertex);
if (forkVertex == null_vertex) {
forkVertex = v;
}
// In local model thread read bases, adjusting i (should stop at 0)
while (i > 0) {
assert(alignMode == AlignMode::LOCAL);
VD newForkVertex = addVertex(sequence[READPOS]);
add_edge(newForkVertex, forkVertex, g_);
VERTEX_ON_PATH(READPOS, newForkVertex);
forkVertex = newForkVertex;
i--;
}
} else if (reachingMove == EndMove) {
assert(forkVertex == null_vertex && u == exitVertex_ && v == null_vertex);
forkVertex = exitVertex_;
if (alignMode == AlignMode::LOCAL) {
// Find the row # we are coming from, walk
// back to there, threading read bases onto
// graph via forkVertex, adjusting i.
const AlignmentColumn* prevCol = alignmentColumnForVertex.at(prevVertex);
int prevRow = ArgMax(prevCol->Score);
while (i > static_cast<int>(prevRow)) {
VD newForkVertex = addVertex(sequence[READPOS]);
add_edge(newForkVertex, forkVertex, g_);
VERTEX_ON_PATH(READPOS, newForkVertex);
forkVertex = newForkVertex;
i--;
}
}
} else if (reachingMove == MatchMove) {
VERTEX_ON_PATH(READPOS, u);
// if there is an extant forkVertex, join it
if (forkVertex != null_vertex) {
add_edge(u, forkVertex, g_);
forkVertex = null_vertex;
}
// add to existing node
curNodeInfo.Reads++;
i--;
} else if (reachingMove == DeleteMove) {
if (forkVertex == null_vertex) {
forkVertex = v;
}
} else if (reachingMove == ExtraMove || reachingMove == MismatchMove) {
// begin a new arc with this read base
VD newForkVertex = addVertex(sequence[READPOS]);
if (forkVertex == null_vertex) {
forkVertex = v;
}
add_edge(newForkVertex, forkVertex, g_);
VERTEX_ON_PATH(READPOS, newForkVertex);
forkVertex = newForkVertex;
i--;
} else {
throw std::runtime_error("unreachable");
}
v = u;
u = prevVertex;
}
startSpanVertex = v;
// if there is an extant forkVertex, join it to enterVertex
if (forkVertex != null_vertex) {
add_edge(enterVertex_, forkVertex, g_);
startSpanVertex = forkVertex;
forkVertex = null_vertex;
}
if (startSpanVertex != exitVertex_) {
tagSpan(startSpanVertex, endSpanVertex);
}
// all filled in?
assert(outputPath == NULL ||
std::find(outputPath->begin(), outputPath->end(), ((size_t)-1)) == outputPath->end());
#undef READPOS
#undef VERTEX_ON_PATH
}
int main( int argc, char* argv[]) {
if(argc != 2) {
std::cerr << "Usage : ./main.out <edge_file>" << std::endl;
exit(1);
}
Network network;
std::ifstream fin(argv[1]);
std::cerr << "Loading input file" << std::endl;
network.LoadFile( fin );
bool is_weighted = network.IsWeighted();
if( ! is_weighted ) {
std::cerr << "All the link weights are 1. Analyze the network as a non-weighted network." << std::endl;
}
std::pair<double,double> fc;
if( is_weighted ) {
std::cerr << "Conducting percolation analysis" << std::endl;
std::ofstream lrp("link_removal_percolation.dat");
lrp << "#fraction weak_link_removal_lcc susceptibility strong_link_removal_lcc susceptibility" << std::endl;
fc = network.AnalyzeLinkRemovalPercolationVariableAccuracy( 0.02, 0.005, lrp );
lrp.flush();
}
std::cerr << "Calculating local clustering coefficients" << std::endl;
network.CalculateLocalCCs();
if( is_weighted ) {
std::cerr << "Calculating overlaps" << std::endl;
network.CalculateOverlaps();
}
std::cerr << "Calculating degree distribution" << std::endl;
std::ofstream dd("degree_distribution.dat");
const auto degree_distribution = network.DegreeDistribution();
for(const auto& f : degree_distribution ) {
dd << f.first << ' ' << f.second << std::endl;
}
dd.flush();
if( is_weighted ) {
std::cerr << "Calculating link weight distribution" << std::endl;
// double edge_weight_bin_size = 1.0;
std::ofstream ewd("edge_weight_distribution.dat");
for(const auto& f : network.EdgeWeightDistributionLogBin() ) {
ewd << f.first << ' ' << f.second << std::endl;
}
ewd.flush();
}
std::map<double, size_t> strength_distribution;
if( is_weighted ) {
std::cerr << "Calculating node strength distribution" << std::endl;
double avg_s = network.AverageEdgeWeight() * network.AverageDegree();
double strength_bin_size = avg_s * 0.01;
std::ofstream sd("strength_distribution.dat");
strength_distribution = network.StrengthDistribution(strength_bin_size);
for(const auto& f :strength_distribution) {
sd << f.first << ' ' << f.second << std::endl;
}
sd.flush();
}
std::cerr << "Calculating c(k)" << std::endl;
std::ofstream cc_d("cc_degree_correlation.dat");
for(const auto& f : network.CC_DegreeCorrelation() ) {
cc_d << f.first << ' ' << f.second << std::endl;
}
cc_d.flush();
if( is_weighted ) {
std::cerr << "Calculating s(k)" << std::endl;
std::ofstream sdc("strength_degree_correlation.dat");
for(const auto& f : network.StrengthDegreeCorrelation() ) {
sdc << f.first << ' ' << f.second << std::endl;
}
sdc.flush();
}
std::cerr << "Calculating k_nn(k)" << std::endl;
std::ofstream ndc("neighbor_degree_correlation.dat");
for(const auto& f : network.NeighborDegreeCorrelation() ) {
ndc << f.first << ' ' << f.second << std::endl;
}
ndc.flush();
if( is_weighted ) {
std::cerr << "Calculating O(w)" << std::endl;
std::ofstream owc("overlap_weight_correlation.dat");
for(const auto& f : network.OverlapWeightCorrelationLogBin() ) {
owc << f.first << ' ' << f.second << std::endl;
}
owc.flush();
}
std::cerr << "Calculating scalar values" << std::endl;
std::ofstream fout("_output.json");
fout << "{" << std::endl;
fout << " \"NumNodes\": " << network.NumNodes() << ',' << std::endl;
fout << " \"NumEdges\": " << network.NumEdges() << ',' << std::endl;
fout << " \"AverageDegree\": " << network.AverageDegree() << ',' << std::endl;
fout << " \"Assortativity\": " << network.PCC_k_knn() << ',' << std::endl;
fout << " \"ArgMax_Pk\": " << ArgMax( degree_distribution ) << ',' << std::endl;
fout << " \"ClusteringCoefficient\": " << network.ClusteringCoefficient() << ',' << std::endl;
fout << " \"PCC_C_k\": " << network.PCC_C_k();
if( is_weighted ) {
fout << ',' << std::endl;
double ave_w = network.AverageEdgeWeight();
fout << " \"AverageEdgeWeight\": " << ave_w << ',' << std::endl;
double ave_k = network.AverageDegree();
fout << " \"AverageStrength\": " << ave_w * ave_k << ',' << std::endl;
double argmax_ps = ArgMax( strength_distribution );
fout << " \"ArgMax_Ps\": " << argmax_ps << ',' << std::endl;
fout << " \"Normalized_ArgMax_Ps\": " << argmax_ps / (ave_w*ave_k) << ',' << std::endl;
fout << " \"PCC_s_k\": " << network.PCC_s_k() << ',' << std::endl;
fout << " \"AverageOverlap\": " << network.AverageOverlap() << ',' << std::endl;
fout << " \"PCC_O_w\": " << network.PCC_O_w() << ',' << std::endl;
fout << " \"Fc_Ascending\": " << fc.first << ',' << std::endl;
fout << " \"Fc_Descending\": " << fc.second << ',' << std::endl;
fout << " \"Delta_Fc\": " << fc.second - fc.first << std::endl;
}
fout << "}" << std::endl;
return 0;
}
size_t ArgMax(const std::vector<TIn> &in, const std::function<TOut(TIn)> f) {
return ArgMax(Map<TIn, TOut>(in, f));
}
