//~~~~~~~
//8/29/2014
//Recursive function
//Keep Spliting until component gets smaller than let's say 300
void splitCmp (int cmpID, vector<string> trVec, map<int, vector<string>> &rcCounts, map<string, vector<string>> &rcID_reverse, map<vector<string>, int> &obsReadsClasses, map<vector<string>, string> &rcID, map<string, map<string, int>> >f, map <string, string> &fa)
{
//Note: Leave single transcript classes "as is"
int tempCount=0;
//cout<<"Component "<<cmpID<<" has size > 10"<<endl;
//sort the current read class in another vector
vector<string> sortedReadClass;
//Adjust read classes in this component and resplit using DFS (or boost)
//1. Get the read classes for this component and save them in a local variable
//It's faster to search for matching read class in this small structure
map<vector<string>, int> cmpObsReadsClasses; //All obs read class that belong to this component only
//for each read class ID
for (const auto &class_ID : rcCounts[cmpID]){
sortedReadClass.clear();
//for each transcript in this read class rcID_reverse[class_ID]-->print the class and the counts
for (const auto &tr : rcID_reverse[class_ID])
{
sortedReadClass.push_back(tr);
//cout<<tr<<"\t";
}
//cout<<"]"<<"\t"<<obsReadsClasses[rcID_reverse[class_ID]]<<endl;
sort(sortedReadClass.begin(), sortedReadClass.end());
cmpObsReadsClasses[sortedReadClass]=obsReadsClasses[rcID_reverse[class_ID]];
}
#if DEBUG
cout<<"Print all initial read classes that belong to component"<<cmpID<<endl;
print(cmpObsReadsClasses);
//exit(7);
#endif
//2.Compress Read Classes
//Delete the last transcript
//for each read class ID
for (const auto &class_ID : rcCounts[cmpID]){
//cout<<"Processing Read Class: "<<endl;
//cout<<"[\t";
sortedReadClass.clear();
//for each transcript in this read class rcID_reverse[class_ID]-->print the class and the counts
for (const auto &tr : rcID_reverse[class_ID])
{
sortedReadClass.push_back(tr);
//cout<<tr<<"\t";
}
//cout<<"]"<<"\t"<<obsReadsClasses[rcID_reverse[class_ID]]<<endl;
sort(sortedReadClass.begin(), sortedReadClass.end());
//if counts < 10 && readclass size is not 1
if(obsReadsClasses[rcID_reverse[class_ID]] < readCountThresh && rcID_reverse[class_ID].size()>1)
{
//cout<<" Count of read class: "<<endl;
//print(rcID_reverse[class_ID]);
//cout<<" is : "<<obsReadsClasses[rcID_reverse[class_ID]]<<endl;
//keep the initial class before removing elements from the end
vector<string> initialReadClass;
for (const auto &tr : rcID_reverse[class_ID])
initialReadClass.push_back(tr);
sort(initialReadClass.begin(), initialReadClass.end());
bool foundClass=false;
while(!foundClass && !rcID_reverse[class_ID].empty())
{
rcID_reverse[class_ID].pop_back();
if(!rcID_reverse[class_ID].empty())
{
//if there is at least one more transcript in this class after removing
//sort again
sortedReadClass.clear();
for (const auto &tr : rcID_reverse[class_ID])
sortedReadClass.push_back(tr);
sort(sortedReadClass.begin(), sortedReadClass.end());
//#if DEBUG
// cout<<"New Read Class "<<endl;
// cout<<" [";
// for (const auto tr : sortedReadClass)
// cout<<tr<<" ";
// cout<<" ]"<<endl;
//#endif
//no need to search inside all read class ... only those read classes that belong to this component
//cmpObsReadsClasses have been created to address the issue above
if ( cmpObsReadsClasses.find(sortedReadClass) != cmpObsReadsClasses.end() ){
//if this read class is found among the given classes
#if DEBUG
cout<<" [";
cout<<"Read class has been found"<<endl;
for (const auto tr : sortedReadClass)
cout<<tr<<" ";
cout<<" ]"<<endl;
#endif
//increment the new founded class
obsReadsClasses[sortedReadClass]+=obsReadsClasses[initialReadClass]; //global read classes
cmpObsReadsClasses[sortedReadClass]+=obsReadsClasses[initialReadClass]; //local read classes
//delete the old class
obsReadsClasses.erase(initialReadClass);
cmpObsReadsClasses.erase(initialReadClass);
//Note: All structures that are related to obsReadsClasses MUST also be updated
//This includes but not limited to: p_value_new2
foundClass=true;
}
}
else
{
//else after poping we got to the end of the stack-->this means no read class was found
//drop the read and write this read to file the read to which was initially mapped that was initally mapped to initialReadClass
/*~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
//unused_reads<<read_name<<"\t mapped to class: [";
cout<<"\t Initial class. (NOT Compatible with any other class): [";
for (const auto tr : initialReadClass)
cout<<tr<<" ";
cout<<" ]"<<endl;
//discarded_reads++;
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~*/
//delete the old class
obsReadsClasses.erase(initialReadClass);
cmpObsReadsClasses.erase(initialReadClass);
#if DEBUG
//cout<<"Read "<<read_name<<" was NOT mapped to any class"<<endl;
cout<<"Initial Class: [";
for (const auto tr : initialReadClass)
cout<<tr<<" ";
cout<<" ]"<<endl;
#endif
}
} //end while class not found
}
}//end: for each read class ID
//cout<<"Print adjusted read classes that belong to this component"<<endl;
//print(cmpObsReadsClasses);
//2. Maybe create again the graph and run boost? (some recursion needs to be done here)
//!!!the entire p_value_new must be updated?
struct Vertex {
std::string vertexName;
};
//typedef adjacency_list <vecS, vecS, undirectedS, Vertex> MyGraph;
typedef boost::labeled_graph<adjacency_list<vecS, vecS, undirectedS, Vertex>, std::string> MyGraph;
//The side effect of this is that one need to use graph() member function to make some algorithms work:
typedef boost::graph_traits<MyGraph>::vertex_descriptor VertexID;
MyGraph G;
//We'll use d_value (p_value) to hold the adjacency matrix
map< string, map< vector<string>, double> > p_value_new2; //this is d_value
//local variable
//for each OBS Read class
for (const auto &read_class: cmpObsReadsClasses){
//for each transcript in the current class
for (const auto &tr: read_class.first){
//if this value exists --> then we have an error
if(p_value_new2[tr][read_class.first]){
cout<<"Error: p value new2 already exists!!!"<<endl;
exit(1);
}
p_value_new2[tr][read_class.first]=read_class.second;//enter the counts for each class
}
}
#if DEBUG
cout<<"\nPrint p_value_new2 (local):"<<endl;
//print(p_value_new2);
//exit(7);
#endif
//Build the graph
//for each transcript --
for (const auto &tr : p_value_new2){
// Create vertices in that graph
VertexID u = boost::add_vertex(tr.first,G);
G[tr.first].vertexName = tr.first;
//for each class in the current transcript
for (const auto &read_class : tr.second ){
//string rClassName = "["; //add [ in order to differentiate the classes that contains only one transcripts from the transcripts itself
string rClassName=rcID[read_class.first]; //get just the read class ID
VertexID v = boost::add_vertex(rClassName,G);
G[rClassName].vertexName = rClassName; //in case vertex already exists then this should overwrite
add_edge(u, v, G);
//allCout<<"[Print 1] Edge btwn vertex "<<tr.first<<" and vertex "<<rClassName<<endl;
}
}
std::vector<int> component(num_vertices(G));
int num = connected_components(G, &component[0]);
//cout << "Total number of components: " << num << endl;
//go thorugh components
//cout<<"component size is: "<<component.size()<<endl;
//CREATE MAP THAT WILL HOLD ALL THE COMPONENTS (transcript names):
map<int, vector<string>> cc;
//CREATE MAP THAT WILL HOLD ALL READ CLASSES AND COUNTS
map<int, vector<string>> rcCountsLocal;
vector<int>::size_type i=0;
Vertex & vertex = G.graph()[i];
//cout<<"Vertex "<<i<<" has name: "<<vertex.vertexName<<endl;
//Notes for each component (# of components == # of vertices but the component id changes only when we have a new component)
for (i = 0; i < component.size(); i++){
//cout<<"Component iterations: "<<i<<endl;
Vertex & vertex = G.graph()[i];
//cout<<"i="<<i<<" Element from component: "<<component[i]<<" is "<<vertex.vertexName<<endl;
if(vertex.vertexName[0] != '[')//if the first character is not [
cc[component[i]].push_back(vertex.vertexName);
else
{ //if the first character it is a "[" then this a read class
rcCountsLocal[component[i]].push_back(vertex.vertexName);
}
}//end: for each component
#if DEBUG
cout<<"Print component map:"<<endl;
for(const auto &cmp : cc)
{
cout<<"component "<<cmp.first<<" - size: "<<cmp.second.size()<<" containts: "<<endl;
for(const auto &elem : cmp.second)
cout<<"\t\t"<<elem<<endl;
}
#endif
//Now print to file
//cout<<"\nWrite observed transcripts names\n\t"<<endl;
ofstream d_stream_new2;
ofstream resultStream;
string resultsFile="../singleTrGenes.txt";
resultStream.open(resultsFile.c_str(),ios::app);//open file in append mode (in order to avoid overwritting)
if(!resultStream){
cout<<"Unable to open " <<resultsFile<<endl;
exit(1);
}
//For each component
for(const auto &cmp : cc)
{//for each component there is a vector with transcript names
//string glblCmpId=dirPrefix+"_"+cmp.first; //global component id
subCmpId++;
//Check the size of the component and if it is 1 then just write the results to file
if( cmp.second.size()< 2 )
{
resultStream<<subCmpId<<"\t"<<cmp.second.size()<<"\t"<<obsReadsClasses[cmp.second];
resultStream<<"\t"<<cmp.second.front()<<"\t1"<<endl;
//cmpID---#tr.---#ObsReads---Transcript names
continue; //continue to the next component
}
//~~~~~~~
//8/28/2014
if(cmp.second.size() > cmpSizeThresh)
{
//Prepare directory prefix
//stringstream ss;
//ss << cmp.first;
//dirPrefix+="_"+ss.str();
//increment read count threshold (required for the next function call)
readCountThresh*=2;
//otherwise we'll get the same read classes (no new colapse will happen)
//Recursive Function
//cout<<"splitCmp: Call the recursive function"<<endl;
splitCmp(cmp.first, cmp.second, rcCountsLocal, rcID_reverse, obsReadsClasses, rcID, gtf, fa);
continue;
}
//~~~~~~~
//Else write to files
write2Files(subCmpId, cmp.second, gtf, fa);
d_stream_new2.clear(); //reuse the same stream (just clear the state flags)
stringstream outFile;
//outFile.str(""); //clear the stringstream
//outFile<<"./"<<dirPrefix<<"_"<<cmp.first<<"/obsRCcounts.txt"; //directory already exists from previous
outFile<<"./"<<subCmpId<<"/obsRCcounts.txt";
string d_file_new2=outFile.str();
d_stream_new2.open(d_file_new2.c_str());
if(!d_stream_new2){
cout<<"Unable to open" <<d_file_new2<<endl;
exit(1);
}
//for each read class ID
for (const auto &class_ID : rcCountsLocal[cmp.first]){
d_stream_new2<<"[\t";
//for each transcript in this read class rcID_reverse[class_ID]-->print the class and the counts
for (const auto &tr : rcID_reverse[class_ID])
{
d_stream_new2<<tr<<"\t";
}
d_stream_new2<<"]"<<"\t"<<obsReadsClasses[rcID_reverse[class_ID]]<<endl;
}//end: for each read class ID
d_stream_new2.close();
//Write to file the number of transcripts in each readClass
d_stream_new2.clear(); //reuse the same stream (just clear the state flags)
outFile.str(""); //clear the stringstream
//outFile<<"./"<<dirPrefix<<"_"<<cmp.first<<"/obsRCsize.txt"; //directory already exists from previous
outFile<<"./"<<subCmpId<<"/obsRCsize.txt";
d_file_new2=outFile.str();
d_stream_new2.open(d_file_new2.c_str());
if(!d_stream_new2){
cout<<"Unable to open" <<d_file_new2<<endl;
exit(1);
}
//Matrix for size of read class
vector<vector<double>> rcSize;
//initialize matrix rcSize with zero
for(int i=0;i<rcCountsLocal[cmp.first].size();i++)
{
vector<double> row; // Create an empty row
for(int j=0;j<rcCountsLocal[cmp.first].size();j++)
{
row.push_back(0.0); // Add an element (column) to the row
}
rcSize.push_back(row); // Add the row to the main vector
}
/*cout<<"print(rcSize)"<<endl;
for(int i=0;i<rcSize.size();i++)
{
for(int j=0;j<rcSize.size();j++)
{
cout<<rcSize[i][j];
}
cout<<endl;
}*/
//fill the diagonal with read classes size.
//for each read class ID
int i=0;
for (const auto &class_ID : rcCountsLocal[cmp.first])
{
rcSize[i][i]=(double)1/rcID_reverse[class_ID].size();
i++;
}
//print Matrix to file
for(int i=0;i<rcCountsLocal[cmp.first].size();i++)
{
for(int j=0;j<rcCountsLocal[cmp.first].size();j++)
{
d_stream_new2<<rcSize[i][j]<<"\t";
}
d_stream_new2<<endl;
}
d_stream_new2.close();
}//end:for each cc
resultStream.close();
//~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
}//end splitCmp
int main(int argc,char *argv[]){
//cout<<"\nRunning "<<argv[0]<<endl;
if(argc<4){
cout<<HELPMESSAGE<<endl;
exit(1);
}
double EPS = 0.01;
int iterations=0;
string obsRBowtie_file=argv[1]; //aligned Observed reads
string gtf_file=argv[2];
string fa_file=argv[3];
clock_t begin = clock(); //used for measuring entire elapsed time for this function
//~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
//~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
cout<<"\n~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"<<endl;
cout<<"["<<current_time()<<"] Load GTF "<<endl;
cout<<"~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"<<endl;
map<string, map<string, int>> gtf;
loadGTF(gtf_file, gtf);
//Write to file transcript lengths //This file is used to replace the use of grep for computing transcript lenght (grep is slow)
ofstream outGlobalTrLength;
string globalTrLengthFile="../trLen.txt";
outGlobalTrLength.open(globalTrLengthFile.c_str());
if(!outGlobalTrLength){
cout<<"Unable to open " <<globalTrLengthFile<<endl;
exit(1);
}
for (const auto &tr : gtf){
outGlobalTrLength<<tr.first<<"\t"<<tr.second.begin()->second<<endl;
}
outGlobalTrLength.close();
cout<<"\n~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"<<endl;
cout<<"["<<current_time()<<"] ~~~~~ Done! ~~~~~"<<endl;
cout<<"~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"<<endl<<endl;
cout<<"\n~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"<<endl;
cout<<"["<<current_time()<<"] Load FA File"<<endl;
cout<<"~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"<<endl;
map<string, string> fa;
loadFA(fa_file, fa);
cout<<"\n~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"<<endl;
cout<<"["<<current_time()<<"] ~~~~~ Done! ~~~~~"<<endl;
cout<<"~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"<<endl<<endl;
cout<<"\n~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"<<endl;
cout<<"["<<current_time()<<"] Compute Observed Reads Classes "<<endl;
cout<<"~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"<<endl;
//Extract OBS reads classes from Bowtie output
map<vector<string>, int> obsReadsClasses;
cout<<"\nMain: Parsing Bowtie file: "<<endl;
extract_obsRC(obsRBowtie_file, obsReadsClasses);
//~~~~~~~~~~~~~~~~~~~~~~~
//08/26/2014~~~~~~~~~~~~~
//delete classes with count 10
//for each read class
//for (const auto &r_class : obsReadsClasses){
// if(r_class.second<10)
// obsReadsClasses.erase(r_class.first);
//}
//~~~~~~~~~~~~~~~~~~~~~~~~~~~
//Create an ID for each Read Class --- this is just a current solution --- needs improvement --- not smart because will double the memory
map<vector<string>, string> rcID;
map<string, vector<string>> rcID_reverse; //in this way we have a bidirectional map (which is stupid anyway...but what to do?)
int total_obs_reads=0;
int id=0;
//for each readClass in obsReadsClasses add an id and Compute Total number of Observed reads
for (const auto &r_class : obsReadsClasses){
stringstream cID_ss;
cID_ss<<"["<<id<<"]";//these square brackets were inserted in order to avoid modifying the code down when it checks for "["
rcID[r_class.first]=cID_ss.str();
rcID_reverse[cID_ss.str()]=r_class.first;
id++;
total_obs_reads+=r_class.second;
}
cout<<"\n~~~ Total number of Observed reads = "<<total_obs_reads<<endl;
cout<<"Total number of Observed Reads classes is: "<<obsReadsClasses.size()<<endl;
#if DEBUG
cout<<"\nOBS Reads Classes \t Size:"<<endl;
//print(obsReadsClasses);
cout<<endl;
//exit(7);
//~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
//This might be only for debugging
//07/30/2014: Print all observed classes to file
ofstream allRCout; // all observed read classes
string allRCFile="../allObsReadClasses.txt";
allRCout.open(allRCFile.c_str(),ios::app);//open file in append mode (in order to avoid overwritting)
if(!allRCout){
cout<<"Unable to open " <<allRCFile<<endl;
exit(1);
}
for (const auto &read_class: obsReadsClasses){ //for each read class
allRCout<<"[ ";
for (const auto &tr : read_class.first)//for each transcript in this class
allRCout<<tr<<" ";
allRCout<<"] "<<read_class.second<<endl;
}
allRCout.close();
//~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
#endif
//cout<<"\n~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"<<endl;
//cout<<"["<<current_time()<<"] ~~~~~ Done! ~~~~~"<<endl;
//cout<<"~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"<<endl<<endl;
//~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
//~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
cout<<"\n\n~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"<<endl;
cout<<"["<<current_time()<<"] Computing adjacency matrix between reads (classes) and transcripts... "<<endl;
cout<<"~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"<<endl;
//We'll use d_value (p_value) to hold the adjacency matrix
map< string, map< vector<string>, double> > p_value_new2; //this is d_value
//for each OBS Read class
for (const auto &read_class: obsReadsClasses){
//for each transcript in the current class
for (const auto &tr: read_class.first){
//if this value exists --> then we have an error
if(p_value_new2[tr][read_class.first]){
cout<<"Error: p value new2 already exists!!!"<<endl;
exit(1);
}
p_value_new2[tr][read_class.first]=read_class.second;//enter the counts for each class
}
}
#if DEBUG
cout<<"\nPrint p_value_new2"<<endl;
//print(p_value_new2);
//exit(7);
#endif
//cout<<"\n~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"<<endl;
//cout<<"["<<current_time()<<"] ~~~~~ Done: adjacency matrix has been computed"<<endl;
//cout<<"~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"<<endl<<endl;
/*~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~*/
cout<<"\n\n~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"<<endl;
cout<<"["<<current_time()<<"] Computing Connected Components ... "<<endl;
cout<<"~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"<<endl;
//Both DFS and BFS complete on O(n) time, using O(n) memory, where n is matrix size. But BFS it doesn't suffer from stack overflow problem, and it doesn't spend time on recursive calls.
//Anyway, let's first use boost library
//*********************************************For debug only*******************************//////////
//08/24/2014: Print all components and edges to file
/*
ofstream allCout; // all components
string allCFile="../allComponents.txt";
allCout.open(allCFile.c_str(),ios::app);//open file in append mode (in order to avoid overwritting)
if(!allCout){
cout<<"Unable to open " <<allCFile<<endl;
exit(1);
}
*/
//*****************************************////////////////////////////////////////////////////////////
struct Vertex {
std::string vertexName;
};
//typedef adjacency_list <vecS, vecS, undirectedS, Vertex> MyGraph;
typedef boost::labeled_graph<adjacency_list<vecS, vecS, undirectedS, Vertex>, std::string> MyGraph;
//The side effect of this is that one need to use graph() member function to make some algorithms work:
typedef boost::graph_traits<MyGraph>::vertex_descriptor VertexID;
MyGraph G;
//for each transcript --
for (const auto &tr : p_value_new2){
// Create vertices in that graph
VertexID u = boost::add_vertex(tr.first,G);
G[tr.first].vertexName = tr.first;
//for each class in the current transcript
for (const auto &read_class : tr.second ){
//string rClassName = "["; //add [ in order to differentiate the classes that contains only one transcripts from the transcripts itself
string rClassName=rcID[read_class.first]; //get just the read class ID
//---Lines below have been replaced by line above
//for each transcript in this class -- create a string with the name of the class
//for (const auto &tr_class : read_class.first)
// rClassName+=tr_class;
//cout<<"Class name is: "<<rClassName<<endl;
VertexID v = boost::add_vertex(rClassName,G);
G[rClassName].vertexName = rClassName; //in case vertex already exists then this should overwrite
add_edge(u, v, G);
//allCout<<"[Print 1] Edge btwn vertex "<<tr.first<<" and vertex "<<rClassName<<endl;
}
}
std::vector<int> component(num_vertices(G));
int num = connected_components(G, &component[0]);
cout << "Total number of components: " << num << endl;
/*
std::vector<int>::size_type j;
for (j = 0; j != component.size(); ++j)
allCout << "Vertex " << j <<" is in component " << component[j]<<endl;
allCout << endl;
cout<<"\nPrint components: "<<endl;
for (const auto &c : component) {
cout<<"Component: "<<c<<endl;
}
*/
//~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
//08/22/2014: Print all components to file
/*
MyGraph::vertex_iterator vertexIt, vertexEnd;
boost::tie(vertexIt, vertexEnd) = vertices(G);
for (; vertexIt != vertexEnd; ++vertexIt){
VertexID vertexID = *vertexIt; // dereference vertexIt, get the ID
Vertex & vertex = G.graph()[vertexID];
//The side effect of boost::labeled_graph is that one need to use graph() member function to make some algorithms work
allCout<<"[Print 2] Vertex name is : "<<vertex.vertexName<<endl;
}
allCout.close();
*/
//cout<<"\n~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"<<endl;
//cout<<"["<<current_time()<<"] ~~~~~ Done: Connected Components had been computed"<<endl;
//cout<<"~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"<<endl<<endl;
//~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
//~~~~~ Write Values to Files ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
//~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
cout<<endl<<"~~~~~~~~~~~~~~~~~~~~~~"<<endl;
cout<<"Prepare Values for SimReg"<<endl;
cout<<"~~~~~~~~~~~~~~~~~~~~~~"<<endl;
cout<<"Please wait, this may take few minutes"<<endl;
cout<<"writing to files . . ."<<endl;
/*********************************
*********************************/
/*
//Write OBS Read Classes Names and Frequencies to File
string read_classes_file2="obsCC.txt";
cout<<"\nWrite read classes to: \n\t"<<read_classes_file2<<endl;
ofstream read_classes_stream;
read_classes_stream.open(read_classes_file2.c_str());
if(!read_classes_stream){
cout<<"Unable to open" <<read_classes_file2<<endl;
exit(1);
}
//for each read class
for (const auto &read_class : obsReadsClasses) {
//print to file the class name
read_classes_stream<<"[ ";
for (const auto &cluster : read_class.first){
read_classes_stream<<cluster<<" ";
}
read_classes_stream<<"]\t"<<new_o[read_class.first]<<endl;
}
read_classes_stream.close();
/*********************************************************************************
**********************************************************************************/
/****************************
* Write transcripts from each component to file *
*****************************/
//go thorugh components
//cout<<"component size is: "<<component.size()<<endl;
//CREATE MAP THAT WILL HOLD ALL THE COMPONENTS (transcript names):
map<int, vector<string>> cc;
//CREATE MAP THAT WILL HOLD ALL READ CLASSES AND COUNTS // ### 9/2/2014: Not sure about the counts? The integer looks like the component ID?
map<int, vector<string>> rcCounts;
vector<int>::size_type i=0;
Vertex & vertex = G.graph()[i];
//cout<<"Vertex "<<i<<" has name: "<<vertex.vertexName<<endl;
//Notes for each component (# of components == # of vertices but the component id changes only when we have a new component)
for (i = 0; i < component.size(); i++){
//cout<<"Component iterations: "<<i<<endl;
Vertex & vertex = G.graph()[i];
//cout<<"i="<<i<<" Element from component: "<<component[i]<<" is "<<vertex.vertexName<<endl;
if(vertex.vertexName[0] != '[')//if the first character is not [
cc[component[i]].push_back(vertex.vertexName);
else
{ //if the first character it is a "[" then this a read class
rcCounts[component[i]].push_back(vertex.vertexName);
}
}//end: for each component
#if DEBUG
cout<<"Print component map:"<<endl;
for(const auto &cmp : cc)
{
cout<<"component "<<cmp.first<<" - size: "<<cmp.second.size()<<" containts: "<<endl;
for(const auto &elem : cmp.second)
cout<<"\t\t"<<elem<<endl;
}
#endif
//Now print to file
cout<<"\nWrite observed transcripts names\n\t"<<endl;
ofstream resultStream;
string resultsFile="../singleTrGenes.txt";
resultStream.open(resultsFile.c_str(),ios::app);//open file in append mode (in order to avoid overwritting)
if(!resultStream){
cout<<"Unable to open " <<resultsFile<<endl;
exit(1);
}
//For each component
for(const auto &cmp : cc)
{//for each component there is a vector with transcript names
//Check the size of the component and if it is 1 then just write the results to file
if( cmp.second.size()< 2 )
{
resultStream<<cmp.first<<"\t"<<cmp.second.size()<<"\t"<<obsReadsClasses[cmp.second];
resultStream<<"\t"<<cmp.second.front()<<"\t1"<<endl;
//cmpID---#tr.---#ObsReads---Transcript names
continue; //continue to the next component
}
//~~~~~~~
//8/28/2014
if(cmp.second.size() > cmpSizeThresh)
{
//Prepare the prefix for sub-components
//stringstream ss;
//ss << cmp.first;
//string dirPrefix=ss.str();
//Recursive Function
cout<<"Call the recursive function"<<endl;
//Reset the thereshold value everytime this function is called
readCountThresh=1;
splitCmp(cmp.first, cmp.second, rcCounts, rcID_reverse, obsReadsClasses, rcID, gtf, fa);
continue;
}
//~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
//#########################################################################
write2Files(cmp.first, cmp.second, gtf, fa);
//#########################################################################
//~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
ofstream d_stream_new2;
//d_stream_new2.clear(); //reuse the same stream (just clear the state flags)
stringstream outFile;
//outFile.str(""); //clear the stringstream
outFile<<"./"<<cmp.first<<"/obsRCcounts.txt"; //directory already exists from previous
string d_file_new2=outFile.str();
d_stream_new2.open(d_file_new2.c_str());
if(!d_stream_new2){
cout<<"Unable to open" <<d_file_new2<<endl;
exit(1);
}
//for each read class ID
for (const auto &class_ID : rcCounts[cmp.first]){
d_stream_new2<<"[\t";
//for each transcript in this read class rcID_reverse[class_ID]-->print the class and the counts
for (const auto &tr : rcID_reverse[class_ID])
{
d_stream_new2<<tr<<"\t";
}
d_stream_new2<<"]"<<"\t"<<obsReadsClasses[rcID_reverse[class_ID]]<<endl;
}//end: for each read class ID
d_stream_new2.close();
//Write to file the number of transcripts in each readClass
d_stream_new2.clear(); //reuse the same stream (just clear the state flags)
outFile.str(""); //clear the stringstream
outFile<<"./"<<cmp.first<<"/obsRCsize.txt"; //directory already exists from previous
d_file_new2=outFile.str();
d_stream_new2.open(d_file_new2.c_str());
if(!d_stream_new2){
cout<<"Unable to open" <<d_file_new2<<endl;
exit(1);
}
//Matrix for size of read class
vector<vector<double>> rcSize;
//initialize matrix rcSize with zero
for(int i=0;i<rcCounts[cmp.first].size();i++)
{
vector<double> row; // Create an empty row
for(int j=0;j<rcCounts[cmp.first].size();j++)
{
row.push_back(0.0); // Add an element (column) to the row
}
rcSize.push_back(row); // Add the row to the main vector
}
/*
cout<<"print(rcSize)"<<endl;
for(int i=0;i<rcSize.size();i++)
{
for(int j=0;j<rcSize.size();j++)
{
cout<<rcSize[i][j];
}
cout<<endl;
}
exit(7);*/
//fill the diagonal with read classes size.
//for each read class ID
int i=0;
for (const auto &class_ID : rcCounts[cmp.first])
{
rcSize[i][i]=(double)1/rcID_reverse[class_ID].size();
i++;
}
//print Matrix to file
for(int i=0;i<rcCounts[cmp.first].size();i++)
{
for(int j=0;j<rcCounts[cmp.first].size();j++)
{
d_stream_new2<<rcSize[i][j]<<"\t";
}
d_stream_new2<<endl;
}
d_stream_new2.close();
}//end:for each cc
resultStream.close();
//~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
clock_t end = clock();
double elapsed_secs = double(end - begin) / CLOCKS_PER_SEC;
cout<<"\nDone - Elapsed time: "<<elapsed_secs<<endl;
}//end main