void SnpEstimation::estimate(GenotypeFileHandler &genotypeFileHandler, boost::ptr_vector<Snp> &snpList, boost::ptr_vector<Region> ®ionList){
// This contains the genotypes from the reference panel, the main container for this function
boost::ptr_deque<Genotype> genotype;
// This contains the index of the SNPs included in the genotype
std::deque<size_t> snpLoc;
// This should be invoked when we came to the end of chromosome / region
bool windowEnd = false;
// Initialize the linkage and decompose
Linkage linkage(m_thread);
Decomposition decompose(m_thread);
fprintf(stderr, "Estimate SNP heritability\n\n");
size_t doneItems = 0;
size_t totalSnp = snpList.size();
std::string chr = "";
// This is use for indicating whether if the whole genome is read
bool completed = false;
std::vector<size_t> boundary;
bool starting = true;
size_t checking = 0; //DEBUG
while(!completed){
// Keep doing this until the whole genome is read
progress(doneItems, totalSnp, chr);
// only used when the finalizeBuff is true, this indicate whether if the last block is coming from somewhere new
bool retainLastBlock=false;
while(boundary.size() < 5 && !completed && !windowEnd){
genotypeFileHandler.getBlock(snpList, genotype, snpLoc, windowEnd, completed,boundary, false);
bool boundChange = false;
linkage.construct(genotype, snpLoc, boundary, snpList, m_ldCorrection, boundChange);
if(boundChange && boundary.back()==snpLoc.size()){
windowEnd=true;
boundary.pop_back();
}
else if(boundChange&&snpList.at(snpLoc[boundary.back()]).getLoc()- snpList.at(snpLoc[boundary.back()-1]).getLoc() > m_blockSize){
retainLastBlock = true;
windowEnd=true;
}
genotypeFileHandler.getBlock(snpList, genotype, snpLoc, windowEnd, completed,boundary, true);
linkage.construct(genotype, snpLoc, boundary, snpList, m_ldCorrection, boundChange);
}
if(windowEnd && !retainLastBlock && boundary.size() > 2){ //Check whether if we need to merge the two blocks
size_t indexOfLastSnpOfSecondLastBlock = snpLoc.at(boundary.back()-1); // just in case
size_t lastSnp = snpLoc.back();
if(snpList.at(lastSnp).getLoc()-snpList.at(indexOfLastSnpOfSecondLastBlock).getLoc() <= m_blockSize){
boundary.pop_back();
bool boundChange=false;
linkage.construct(genotype, snpLoc, boundary, snpList, m_ldCorrection, boundChange);
}
}
if(retainLastBlock && !windowEnd) throw std::runtime_error("Impossible combination of windowEnd and retain last block!");
decompose.run(linkage, snpLoc, boundary, snpList, windowEnd, !retainLastBlock, starting, regionList);
doneItems= snpLoc.at(boundary.back());
chr = snpList[snpLoc[boundary.back()]].getChr();
if(retainLastBlock){
// Then we must remove everything except the last block
// because finalizeBuff must be true here
size_t update = boundary.back();
genotype.erase(genotype.begin(), genotype.begin()+update);
snpLoc.erase(snpLoc.begin(), snpLoc.begin()+update);
boundary.clear();
boundary.push_back(0);
linkage.clear(update);
starting = true;
windowEnd = false;
retainLastBlock = false;
}
else if(windowEnd){
snpLoc.clear();
boundary.clear();
genotype.clear();
linkage.clear();
starting = true;
windowEnd = false;
retainLastBlock = false;
}
else{
starting = false;
size_t update=boundary[1];
genotype.erase(genotype.begin(), genotype.begin()+update);
snpLoc.erase(snpLoc.begin(), snpLoc.begin()+update);
linkage.clear(update);
for(size_t i = 0; i < boundary.size()-1; ++i) boundary[i] = boundary[i+1]-update;
boundary.pop_back();
}
}
progress(totalSnp, totalSnp, "");
fprintf(stderr, "\n\nEstimated the SNP Heritability, now proceed to output\n");
}
