int main(int argc, char* argv[])
{
size_t kmerSize = 7;
const char* sequences[] = {
"AGGCGCTAGGGAGAGGATGATGAAA",
"AGGCGCTCGGGAGAGGATGATGAAA",
"AGGCGCTTGGGAGAGGATGATGAAA"
};
try
{
// Create the graph
Graph graph = Graph::create (
new BankStrings (sequences, ARRAY_SIZE(sequences)),
"-kmer-size %d -abundance-min 1 -verbose 0", kmerSize
);
// Take the first node, which should be a branching node
Node node = graph.buildNode (sequences[0]);
// Retrieve the branching neighbors for the node (as BRANCHING EDGES!)
Graph::Vector<BranchingEdge> branchingNeighbors =
graph.successors<BranchingEdge> (node);
std::cout
<< "We found "
<< branchingNeighbors.size()
<< " branching neighbors from node "
<< graph.toString(node)
<< std::endl;
// Iterate over the branching neighbors to print them
for (size_t i = 0; i < branchingNeighbors.size(); i++)
{
// Note; we don't display all the transition nucleotides, only the
// first transition nucleotide. We also display the number of
// transitions needed to link the two branching nodes.
std::cout << graph.toString (branchingNeighbors[i]) << std::endl;
}
}
catch(Exception& e)
{
std::cerr << "EXCEPTION: " << e.getMessage() << std::endl;
}
return EXIT_SUCCESS;
}
int main (int argc, char* argv[])
{
/** We create a command line parser. */
OptionsParser parser;
parser.push_back (new OptionOneParam (STR_URI_INPUT, "graph file", true));
IProperties* params = 0;
try {
/** We parse the user options. */
params = parser.parse (argc, argv);
}
catch (OptionFailure& e)
{
e.getParser().displayErrors (stdout);
e.getParser().displayHelp (stdout);
return EXIT_FAILURE;
}
// We create the graph with the bank and other options
Graph graph = Graph::load (params->getStr(STR_URI_INPUT));
// We create a graph marker.
GraphMarker<BranchingNode> marker (graph);
// We create an object for Breadth First Search for the de Bruijn graph.
BFS<BranchingNode> bfs (graph);
// We want to compute the distribution of connected components of the branching nodes.
// - key is a connected component class (for a given number of branching nodes for this component)
// - value is the number of times this component class occurs in the branching sub graph
map<size_t,Entry> distrib;
// We get an iterator for all nodes of the graph. We use a progress iterator to get some progress feedback
ProgressGraphIterator<BranchingNode,ProgressTimer> itBranching (graph.iterator<BranchingNode>(), "statistics");
// We want to know the number of connected components
size_t nbConnectedComponents = 0;
// We define some kind of unique identifier for a couple (indegree,outdegree)
map <InOut_t, size_t> topology;
size_t simplePathSizeMin = ~0;
size_t simplePathSizeMax = 0;
// We want time duration of the iteration
TimeInfo ti;
ti.start ("compute");
// We loop the branching nodes
for (itBranching.first(); !itBranching.isDone(); itBranching.next())
{
// We get branching nodes neighbors for the current branching node.
Graph::Vector<BranchingEdge> successors = graph.successors <BranchingEdge> (*itBranching);
Graph::Vector<BranchingEdge> predecessors = graph.predecessors<BranchingEdge> (*itBranching);
// We increase the occurrences number for the current couple (in/out) neighbors
topology [make_pair(predecessors.size(), successors.size())] ++;
// We loop the in/out neighbors and update min/max simple path size
for (size_t i=0; i<successors.size(); i++)
{
simplePathSizeMax = std::max (simplePathSizeMax, successors[i].distance);
simplePathSizeMin = std::min (simplePathSizeMin, successors[i].distance);
}
for (size_t i=0; i<predecessors.size(); i++)
{
simplePathSizeMax = std::max (simplePathSizeMax, predecessors[i].distance);
simplePathSizeMin = std::min (simplePathSizeMin, predecessors[i].distance);
}
// We skip already visited nodes.
if (marker.isMarked (*itBranching)) {
continue;
}
// We launch the breadth first search; we get as a result the set of branching nodes in this component
const set<BranchingNode>& component = bfs.run (*itBranching);
// We mark the nodes for this connected component
marker.mark (component);
// We update our distribution
distrib[component.size()].nbOccurs += 1;
// We update the number of connected components.
nbConnectedComponents++;
}
ti.stop ("compute");
// We compute the total number of branching nodes in all connected components.
size_t sumOccurs = 0;
size_t sumKmers = 0;
for (map<size_t,Entry>::iterator it = distrib.begin(); it != distrib.end(); it++)
{
sumOccurs += it->first*it->second.nbOccurs;
sumKmers += it->second.nbKmers;
}
// We sort the statistics by decreasing occurrence numbers. Since map have its own ordering, we need to put all
// the data into a vector and sort it with our own sorting criteria.
vector < pair<InOut_t,size_t> > stats;
for (map <InOut_t, size_t>::iterator it = topology.begin(); it != topology.end(); it++) {
stats.push_back (*it);
}
sort (stats.begin(), stats.end(), CompareFct);
// Note: it must be equal to the number of branching nodes of the graph
assert (sumOccurs == itBranching.size());
// We aggregate the computed information
Properties props ("topology");
props.add (1, "graph");
props.add (2, "name", "%s", graph.getName().c_str());
props.add (2, "db_input", "%s", graph.getInfo().getStr("input").c_str());
props.add (2, "db_nb_seq", "%d", graph.getInfo().getInt("sequences_number"));
props.add (2, "db_size", "%d", graph.getInfo().getInt("sequences_size"));
props.add (2, "kmer_size", "%d", graph.getInfo().getInt("kmer_size"));
props.add (2, "kmer_nks", "%d", graph.getInfo().getInt("nks"));
props.add (2, "nb_nodes", "%d", graph.getInfo().getInt("kmers_nb_solid"));
props.add (2, "nb_branching_nodes", "%d", graph.getInfo().getInt("nb_branching"));
props.add (2, "percent_branching_nodes", "%.1f",
graph.getInfo().getInt("kmers_nb_solid") > 0 ?
100.0 * (float)graph.getInfo().getInt("nb_branching") / (float) graph.getInfo().getInt("kmers_nb_solid") : 0
);
props.add (1, "branching_nodes");
props.add (2, "simple_path");
props.add (3, "size_min", "%d", simplePathSizeMin);
props.add (3, "size_max", "%d", simplePathSizeMax);
props.add (2, "neighborhoods");
for (size_t i=0; i<stats.size(); i++)
{
props.add (3, "neighborhood", "in=%d out=%d", stats[i].first.first, stats[i].first.second);
props.add (4, "nb_bnodes", "%d", stats[i].second);
props.add (4, "percentage", "%5.2f", itBranching.size() > 0 ?
100.0*(float)stats[i].second / (float)itBranching.size() : 0
);
}
props.add (2, "connected_components");
props.add (3, "nb_classes", "%d", distrib.size());
props.add (3, "nb_components", "%d", nbConnectedComponents);
for (map<size_t,Entry>::iterator it = distrib.begin(); it!=distrib.end(); it++)
{
props.add (3, "component_class");
props.add (4, "nb_occurs", "%d", it->second.nbOccurs);
props.add (4, "nb_bnodes", "%d", it->first);
props.add (4, "freq_bnodes", "%f", sumOccurs > 0 ?
100.0*(float)(it->first*it->second.nbOccurs) / (float)sumOccurs : 0
);
}
props.add (1, ti.getProperties("time"));
// We dump the results in a XML file in the current directory
XmlDumpPropertiesVisitor v (graph.getName() + ".xml", false);
props.accept (&v);
return EXIT_SUCCESS;
}
int main (int argc, char* argv[])
{
/** We create a command line parser. */
OptionsParser parser ("GraphStats");
parser.push_back (new OptionOneParam (STR_URI_GRAPH, "graph input", true));
parser.push_back (new OptionOneParam (STR_NB_CORES, "nb cores", false, "0"));
try
{
/** We parse the user options. */
IProperties* options = parser.parse (argc, argv);
// We load the graph
Graph graph = Graph::load (options->getStr(STR_URI_GRAPH));
// We set the number of cores to be used. Use all available cores if set to 0.
size_t nbCores = options->getInt(STR_NB_CORES);
// We get an iterator for branching nodes of the graph.
// We use a progress iterator to get some progress feedback
ProgressGraphIterator<BranchingNode,ProgressTimer> itBranching (graph.iterator<BranchingNode>(), "statistics");
// We define some kind of unique identifier for a couple (indegree,outdegree)
typedef pair<size_t,size_t> InOut_t;
// We want to gather some statistics during the iteration.
// Note the use of ThreadObject: this object will be cloned N times (one object per thread) and each clone will
// be reachable within the iteration block through ThreadObject::operator()
ThreadObject <map <InOut_t, size_t> > topology;
// We dispatch the iteration on several cores. Note the usage of lambda expression here.
IDispatcher::Status status = Dispatcher(nbCores).iterate (itBranching, [&] (const BranchingNode& node)
{
// We retrieve the current instance of map <InOut_t,size_t> for the current running thread.
map <InOut_t,size_t>& localTopology = topology();
// We get branching nodes neighbors for the current branching node.
Graph::Vector<BranchingEdge> successors = graph.successors <BranchingEdge> (node);
Graph::Vector<BranchingEdge> predecessors = graph.predecessors<BranchingEdge> (node);
// We increase the occurrences number for the current couple (in/out) neighbors
localTopology [make_pair(predecessors.size(), successors.size())] ++;
});
// Now, the parallel processing is done. We want now to aggregate the information retrieved
// in each thread in a single map.
// We get each map<InOut_t,size_t> object filled in each thread, and we add its data into the "global" map.
// The global map is reachable through the ThreadObject::operator*. The "topology.foreach" will loop over
// all cloned object used in the threads.
topology.foreach ([&] (const map <InOut_t, size_t>& t)
{
// We update the occurrence of the current couple (in/out)
for_each (t.begin(), t.end(), [&] (const pair<InOut_t, size_t>& p) { (*topology)[p.first] += p.second; });
});
// We sort the statistics by decreasing occurrence numbers. Since map have its own ordering, we need to put all
// the data into a vector and sort it with our own sorting criteria.
vector < pair<InOut_t,size_t> > stats;
for (auto it = topology->begin(); it != topology->end(); it++) { stats.push_back (*it); }
sort (stats.begin(), stats.end(), [=] (const pair<InOut_t,size_t>& a, const pair<InOut_t,size_t>& b) { return a.second > b.second; });
printf ("\nThere are %d branching nodes with the following distribution: \n", itBranching.size());
size_t sum=0;
for (size_t i=0; i<stats.size(); i++)
{
sum += stats[i].second;
printf (" [in=%d out=%d] nb=%7d percent=%5.2f distrib=%5.2f\n",
stats[i].first.first,
stats[i].first.second,
stats[i].second,
100.0*(float)stats[i].second / (float)itBranching.size(),
100.0*(float)sum / (float)itBranching.size()
);
}
printf ("\nDone on %d cores in %.2f sec\n\n", status.nbCores, (float)status.time/1000.0);
}
catch (OptionFailure& e)
{
return e.displayErrors (std::cout);
}
catch (Exception& e)
{
std::cerr << "EXCEPTION: " << e.getMessage() << std::endl;
}
return EXIT_SUCCESS;
}