Dsuite_utils.cpp

//
//  Dsuite_utils.cpp
//  Dsuite
//
//  Created by Milan Malinsky on 02/04/2019.
//

#include "Dsuite_utils.h"

long double normalCDF(double x) // Phi(-∞, x) aka N(x)
{
    return erfcl(-x/std::sqrt(2))/2;
}

double Fd_Denom_perVariant(double p1, double p2, double p3, double pO) {
    double Fd_Denom = 0;
    if (p2 > p3) Fd_Denom = ((1-p1)*p2*p2*(1-pO)) - (p1*(1-p2)*p2*(1-pO));
    else Fd_Denom = ((1-p1)*p3*p3*(1-pO)) - (p1*(1-p3)*p3*(1-pO));
    return Fd_Denom;
}

double fG_Denom_perVariant(double p1, double p3a, double p3b, double pO) {
    double fG_Denom = ((1-p1)*p3a*p3b*(1-pO)) - (p1*(1-p3a)*p3b*(1-pO));
    return fG_Denom;
}

double FdM_Denom_perVariant(double p1, double p2, double p3, double pO) {
    double FdM_Denom = 0;
    if (p1 <= p2) {
        if (p2 > p3) FdM_Denom = ((1-p1) * p2 * p2 * (1-pO)) - (p1 * (1-p2) * p2 * (1-pO));
        else FdM_Denom = ((1-p1) * p3 * p3 * (1-pO)) - (p1 * (1-p3) * p3 * (1-pO));
    } else {
        if (p1 > p3) FdM_Denom = -(((1-p1)*p2*p1*(1-pO)) - (p1*(1-p2)*p1*(1-pO)));
        else FdM_Denom = -(((1-p3)*p2*p3*(1-pO)) - (p3*(1-p2)*p3*(1-pO)));
    }
    return FdM_Denom;
}



// Works only on biallelic markers
void GeneralSetCounts::getSetVariantCounts(const std::vector<std::string>& genotypes, const std::map<size_t, string>& posToSpeciesMap) {
    
    getBasicCounts(genotypes, posToSpeciesMap);
    
    // If at least one of the outgroup individuals has non-missing data
    // Find out what is the "ancestral allele" - i.e. the one more common in the outgroup
    try {
        if (setAlleleCounts.at("Outgroup") > 0) {
            if ((double)setAltCounts.at("Outgroup")/setAlleleCounts.at("Outgroup") < 0.5) { AAint = AncestralAlleleRef; }
            else { AAint = AncestralAlleleAlt; }
        }
    } catch (std::out_of_range& e) { AAint = AncestralAlleleMissing; }
    
    // Now fill in the allele frequencies
    double totalAAF = 0; double totalDAF = 0; int numNonZeroCounts = 0;
    for(std::map<string,int>::iterator it = setAltCounts.begin(); it != setAltCounts.end(); ++it) {
        if (setAlleleCounts.at(it->first) > 0) {
            numNonZeroCounts++;
            double thisAAF = (double)setAltCounts.at(it->first)/setAlleleCounts.at(it->first);
            setAAFs[it->first] = thisAAF; totalAAF += thisAAF;
            if (AAint == 0) { // Ancestral allele seems to be the ref, so derived is alt
                setDAFs[it->first] = thisAAF; totalDAF += thisAAF;
            } else if (AAint == 1) { // Ancestral allele seems to be alt, so derived is ref
                setDAFs[it->first] = (1 - thisAAF); totalDAF += (1 - thisAAF);
            }
        }
    }
    averageAAF = totalAAF/numNonZeroCounts; averageDAF = totalDAF/numNonZeroCounts;
}

int GeneralSetCounts::returnFormatTagPosition(std::vector<std::string>& format, const std::string& tag) {
    // Find the position of GQ (genotype quality) in the genotypeData vector below
    std::vector<std::string>::iterator TAGit; int TAGi = std::numeric_limits<int>::min();
    TAGit = find (format.begin(), format.end(), tag);
    if (TAGit == format.end()) {
        // std::cerr << "This variant hasn't got associated per-sample GQ info" << std::endl;
    } else {
        TAGi = (int)std::distance( format.begin(), TAGit );
        //hasGQ = true;
    }
    return TAGi;
}


int GeneralSetCounts::checkForGenotypeLikelihoodsOrProbabilities(const std::vector<std::string>& vcfLineFields) {
    std::vector<std::string> format = split(vcfLineFields[8], ':');
    if (format.size() == 1) return LikelihoodsProbabilitiesAbsent; // The GT tag must be present in the first place
    
    int likelihoodsOrProbabilitiesTagPosition = returnFormatTagPosition(format, "GP");
    if (likelihoodsOrProbabilitiesTagPosition != std::numeric_limits<int>::min()) { likelihoodsProbabilitiesType = LikelihoodsProbabilitiesGP; }
    else {
        likelihoodsOrProbabilitiesTagPosition = returnFormatTagPosition(format, "GL");
        if (likelihoodsOrProbabilitiesTagPosition != std::numeric_limits<int>::min()) { likelihoodsProbabilitiesType = LikelihoodsProbabilitiesGL; }
        else {
            likelihoodsOrProbabilitiesTagPosition = returnFormatTagPosition(format, "PL");
            if (likelihoodsOrProbabilitiesTagPosition != std::numeric_limits<int>::min()) { likelihoodsProbabilitiesType = LikelihoodsProbabilitiesPL; }
        }
    }
    return likelihoodsOrProbabilitiesTagPosition;
}

double getExpectedGenotype(const std::vector<double>& thisProbabilities) {
    double Egenotype = thisProbabilities[1] + 2*thisProbabilities[2];
    return Egenotype;
}

void transformFromPhred(std::vector<double>& thisLikelihoods) {

    thisLikelihoods[0] = pow(10,-(thisLikelihoods[0]/10.0));
    thisLikelihoods[1] = pow(10,-(thisLikelihoods[1]/10.0));
    thisLikelihoods[2] = pow(10,-(thisLikelihoods[2]/10.0));
}

void transformFromGL(std::vector<double>& thisLikelihoods) {

    thisLikelihoods[0] = pow(10,(thisLikelihoods[0]/10.0));
    thisLikelihoods[1] = pow(10,(thisLikelihoods[1]/10.0));
    thisLikelihoods[2] = pow(10,(thisLikelihoods[2]/10.0));
}

std::vector<double> GeneralSetCounts::probabilitiesFromLikelihoods(const std::vector<double>& thisLikelihoods, const string& species) {
    std::vector<double> thisProbabilities; thisProbabilities.assign(3, 0.0);
    double multiple0 = thisLikelihoods[0]*setHWEpriorsFromAAFfromGT[species][0];
    double multiple1 = thisLikelihoods[1]*setHWEpriorsFromAAFfromGT[species][1];
    double multiple2 = thisLikelihoods[2]*setHWEpriorsFromAAFfromGT[species][2];
    double sum = multiple0 + multiple1 + multiple2;
    
    thisProbabilities[0] = multiple0/sum;
    thisProbabilities[1] = multiple1/sum;
    thisProbabilities[2] = multiple2/sum;
    
    return thisProbabilities;
}
 
void GeneralSetCounts::setHWEpriorsFromAFfromGT() {
    double AF;
    // Alternative allele frequencies
    for(std::map<string,double>::iterator it = setAAFs.begin(); it != setAAFs.end(); ++it) {
        if (it->second >= 0) AF = it->second; else AF = averageAAF; // This should be average of AFs across populations where it is known
        setHWEpriorsFromAAFfromGT[it->first][0] = pow((1-AF),2);
        setHWEpriorsFromAAFfromGT[it->first][1] = AF*(1-AF);
        setHWEpriorsFromAAFfromGT[it->first][2] = pow(AF,2);
    }
    // Derived allele frequencies
    for(std::map<string,double>::iterator it = setDAFs.begin(); it != setDAFs.end(); ++it) {
        if (it->second >= 0) AF = it->second; else AF = averageDAF; // This should be average of AFs across populations
        setHWEpriorsFromDAFfromGT[it->first][0] = pow((1-AF),2);
        setHWEpriorsFromDAFfromGT[it->first][1] = AF*(1-AF);
        setHWEpriorsFromDAFfromGT[it->first][2] = pow(AF,2);
    }
} 




void GeneralSetCounts::getAFsFromGenotypeLikelihoodsOrProbabilities(const std::vector<std::string>& genotypeFields, const std::map<size_t, string>& posToSpeciesMap, const int likelihoodsOrProbabilitiesTagPosition) {
    if (likelihoodsProbabilitiesType == LikelihoodsProbabilitiesPL || likelihoodsProbabilitiesType == LikelihoodsProbabilitiesGL) {
        setHWEpriorsFromAFfromGT();
    }
    
    for (std::vector<std::string>::size_type i = 0; i < genotypeFields.size(); i++) {
        std::string species; try { species = posToSpeciesMap.at(i); } catch (const std::out_of_range& oor) {
            continue;
        }
       // std::cerr << genotypeFields[i] << std::endl;
        std::string thisLikelihoodsOrProbabilitiesString = split(genotypeFields[i], ':')[likelihoodsOrProbabilitiesTagPosition];
        if (thisLikelihoodsOrProbabilitiesString == ".") continue;
        
        else {
            setAlleleProbCounts.at(species) += 2;
            std::vector<double> thisLikelihoodsOrProbabilities = splitToDouble(thisLikelihoodsOrProbabilitiesString,',');
            std::vector<double> thisProbabilities;
            switch (likelihoodsProbabilitiesType)
            {
                case LikelihoodsProbabilitiesPL:
                    transformFromPhred(thisLikelihoodsOrProbabilities);
                   // print_vector(thisLikelihoodsOrProbabilities, std::cerr);
                    thisProbabilities = probabilitiesFromLikelihoods(thisLikelihoodsOrProbabilities,species);
                    break;
                case LikelihoodsProbabilitiesGL: transformFromGL(thisLikelihoodsOrProbabilities);
                    thisProbabilities = probabilitiesFromLikelihoods(thisLikelihoodsOrProbabilities,species);
                    break;
                case LikelihoodsProbabilitiesGP:
                    thisProbabilities = thisLikelihoodsOrProbabilities;
                    break;
            }
            if (setAAFsFromLikelihoods.at(species) == -1) setAAFsFromLikelihoods.at(species) = 0;
            setAAFsFromLikelihoods.at(species) += getExpectedGenotype(thisProbabilities);
        }
    }
    
    for(std::map<string,double>::iterator it = setAAFsFromLikelihoods.begin(); it != setAAFsFromLikelihoods.end(); ++it) {
        if (setAAFsFromLikelihoods.at(it->first) != -1) {
            double AF = it->second/setAlleleProbCounts.at(it->first);
            it->second = AF;
            if (AAint == AncestralAlleleRef) {
                setDAFsFromLikelihoods.at(it->first) = AF;
            } else if (AAint == AncestralAlleleAlt) {
                setDAFsFromLikelihoods.at(it->first) = (1 - AF);
            }
        }
    }
     
}

void GeneralSetCounts::getAFsFromADtag(const std::vector<std::string>& genotypeFields, const std::map<string, std::vector<size_t>>& setsToPosMap, const int ADTagPosition, const int minDepth) {
    for (std::vector<std::string>::size_type i = 0; i < genotypeFields.size(); i++) {
          // std::cerr << genotypeFields[i] << std::endl;
           std::string thisADstring = split(genotypeFields[i], ':')[ADTagPosition];
           if (thisADstring == ".") {
               std::cerr << "The AD tag info appears to be missing: " << thisADstring << " ; Exiting ..." << std::endl;
               exit(1);
           }
           
           else {
               std::vector<double> ADs = splitToDouble(thisADstring,',');
               if (ADs.size() != 2) {
                   std::cerr << "This AD tag appears malformed: " << thisADstring << " ; Exiting ..." << std::endl;
                   exit(1);
               }
               
               int overallDepth = ADs[0] + ADs[1];
               if (overallDepth >= minDepth) {
                    individualPoolAAFs[i] = ADs[0]/(overallDepth);
               }
           }
       }
       
       for(std::map<string, std::vector<size_t>>::const_iterator it = setsToPosMap.begin(); it != setsToPosMap.end(); ++it) {
           int individualsInThisSet = (int) it->second.size();
           assert(individualsInThisSet > 0);
           if (individualsInThisSet == 1) {
               int pos = (int) it->second[0];
               setPoolAAFs.at(it->first) = individualPoolAAFs[pos];
           } else {
               std::vector<double> thisSetAFs;
               for (int i = 0; i < individualsInThisSet; i++) {
                   int pos = (int) it->second[i];
                   if (individualPoolAAFs[pos] != -1.0) thisSetAFs.push_back(individualPoolAAFs[pos]);
               }
               setPoolAAFs.at(it->first) = vector_average(thisSetAFs);
               
           }
           
           
           if (AAint == AncestralAlleleRef) {
               setPoolDAFs.at(it->first) = setPoolAAFs.at(it->first);
           } else if (AAint == AncestralAlleleAlt && setPoolAAFs.at(it->first) != -1.0) {
               setPoolDAFs.at(it->first) = (1 - setPoolAAFs.at(it->first));
           }
           
               
       }
}


void GeneralSetCountsWithSplits::getAFsFromADtagWithSplits(const std::vector<std::string>& genotypeFields, const std::map<string, std::vector<size_t>>& setsToPosMap, const int ADTagPosition, const int minDepth) {
    
    
    for (std::vector<std::string>::size_type i = 0; i < genotypeFields.size(); i++) {
       // std::cerr << genotypeFields[i] << std::endl;
        std::string thisADstring = split(genotypeFields[i], ':')[ADTagPosition];
        if (thisADstring == ".") {
            std::cerr << "The AD tag info appears to be missing: " << thisADstring << " ; Exiting ..." << std::endl;
            exit(1);
        }
        
        else {
            std::vector<double> ADs = splitToDouble(thisADstring,',');
            if (ADs.size() != 2) {
                std::cerr << "This AD tag appears malformed: " << thisADstring << " ; Exiting ..." << std::endl;
                exit(1);
            }
            
            int overallDepth = ADs[0] + ADs[1];
            if (overallDepth >= minDepth) {
                 individualPoolAAFs[i] = ADs[0]/(overallDepth);
            }
        }
    }
    
    for(std::map<string, std::vector<size_t>>::const_iterator it = setsToPosMap.begin(); it != setsToPosMap.end(); ++it) {
        int individualsInThisSet = (int) it->second.size();
        assert(individualsInThisSet > 0);
        if (individualsInThisSet == 1) {
            int pos = (int) it->second[0];
            setPoolAAFs.at(it->first) = individualPoolAAFs[pos];
            setPoolAAFsplit1.at(it->first) = individualPoolAAFs[pos];
            setPoolAAFsplit2.at(it->first) = individualPoolAAFs[pos];
        } else {
            std::vector<double> thisSetAFs;
            for (int i = 0; i < individualsInThisSet; i++) {
                int pos = (int) it->second[i];
                thisSetAFs.push_back(individualPoolAAFs[pos]);
            }
            setPoolAAFs.at(it->first) = vector_average(thisSetAFs);
            
            // Take care of the splits by random sampling with replacement:
            std::random_device rd;     // only used once to initialise (seed) engine
            std::mt19937 rng(rd());    // random-number engine used (Mersenne-Twister in this case)
            std::uniform_int_distribution<int> uni(0,(individualsInThisSet - 1)); // guaranteed unbiased
            
            std::vector<double> thisSetAFsplit1; std::vector<double> thisSetAFsplit2;
            for (int i = 0; i < individualsInThisSet; i++) {
                int random_pos_s1 = uni(rng);
                int random_pos_s2 = uni(rng);
                thisSetAFsplit1.push_back(individualPoolAAFs[random_pos_s1]);
                thisSetAFsplit2.push_back(individualPoolAAFs[random_pos_s2]);
            }
            setPoolAAFsplit1.at(it->first) = vector_average(thisSetAFsplit1);
            setPoolAAFsplit2.at(it->first) = vector_average(thisSetAFsplit2);
            
        }
        
        if (AAint == AncestralAlleleRef) {
            setPoolDAFs.at(it->first) = setPoolAAFs.at(it->first);
            setPoolDAFsplit1.at(it->first) = setPoolAAFsplit1.at(it->first);
            setPoolDAFsplit2.at(it->first) = setPoolAAFsplit2.at(it->first);
        } else if (AAint == AncestralAlleleAlt && setPoolAAFs.at(it->first) != -1.0) {
            setPoolDAFs.at(it->first) = (1 - setPoolAAFs.at(it->first));
            setPoolDAFsplit1.at(it->first) = (1 - setPoolAAFsplit1.at(it->first));
            setPoolDAFsplit2.at(it->first) = (1 - setPoolAAFsplit2.at(it->first));
        }
            
    }
}


// Only works for diploids for now!!!
void GeneralSetCountsWithSplits::getAFsFromGenotypeLikelihoodsOrProbabilitiesWithSplits(const std::vector<std::string>& genotypeFields, const std::map<size_t, string>& posToSpeciesMap, const int likelihoodsOrProbabilitiesTagPosition, const int pos) {
   
    
    if (likelihoodsProbabilitiesType == LikelihoodsProbabilitiesPL || likelihoodsProbabilitiesType == LikelihoodsProbabilitiesGL) {
        setHWEpriorsFromAFfromGT();
    }
    
    getBasicCountsFromLikelihoodsOrProbabilities(genotypeFields, posToSpeciesMap, likelihoodsOrProbabilitiesTagPosition);
    
     
    // Now fill in the allele frequencies
    for(std::map<string,std::vector<double>>::iterator it = setIndividualExpectedGenotypes.begin(); it != setIndividualExpectedGenotypes.end(); ++it) {
        if (it->first == "") {
            std::cerr << "it->first " << it->first << "\t"; print_vector(it->second, std::cerr); std::cerr << std::endl;
        }
        std::vector<double> thisSetExpectedGenotypes = it->second;
        
        
        if (thisSetExpectedGenotypes.size() > 0) {
            double thisAAF = (double)vector_sum(thisSetExpectedGenotypes)/(2*thisSetExpectedGenotypes.size());
          /* Debug stuff
           if(pos == 1180 || pos == 1046) {
                std::cerr << "pos: " << pos << std::endl;
                std::cerr << "it->first: " << it->first << std::endl;
                print_vector(thisSetExpectedGenotypes, std::cerr);
                std::cerr << "thisAAF: " << thisAAF << std::endl;
            }
           */
            //std::cerr << "species: " << it->first << std::endl;
            // print_vector(thisSetExpectedGenotypes, std::cerr);
            // std::cerr << "thisAAF: " << thisAAF << std::endl;
            setAAFsFromLikelihoods.at(it->first) = thisAAF;
            
            // Take care of the splits by random sampling with replacement:
            std::random_device rd;     // only used once to initialise (seed) engine
            std::mt19937 rng(rd());    // random-number engine used (Mersenne-Twister in this case)
            std::uniform_int_distribution<int> uniAFs(0,((int)thisSetExpectedGenotypes.size() - 1)); // guaranteed unbiased
            
            
            std::vector<double> thisSetIndividualExpectedGenotypesSampledSplit1;
            std::vector<double> thisSetIndividualExpectedGenotypesSampledSplit2;
            for (int i = 0; i < thisSetExpectedGenotypes.size(); i++) {
                int random_pos_s1 = uniAFs(rng);
                int random_pos_s2 = uniAFs(rng);
                thisSetIndividualExpectedGenotypesSampledSplit1.push_back(thisSetExpectedGenotypes[random_pos_s1]);
                thisSetIndividualExpectedGenotypesSampledSplit2.push_back(thisSetExpectedGenotypes[random_pos_s2]);
            }
            
            double thisAAFsplit1 = (double)vector_sum(thisSetIndividualExpectedGenotypesSampledSplit1)/(2*thisSetExpectedGenotypes.size());
           // std::cerr << "thisAAFsplit1: " << thisAAFsplit1 << std::endl;
            double thisAAFsplit2 = (double)vector_sum(thisSetIndividualExpectedGenotypesSampledSplit2)/(2*thisSetExpectedGenotypes.size());
           // std::cerr << "thisAAFsplit2: " << thisAAFsplit2 << std::endl;

            
           // std::cerr << "it->first " << it->first << std::endl;
            try {
            setAAFsplit1fromLikelihoods.at(it->first) = thisAAFsplit1; setAAFsplit2fromLikelihoods.at(it->first) = thisAAFsplit2;
                
            if (AAint == AncestralAlleleRef) { // Ancestral allele seems to be the ref, so derived is alt
                setDAFsFromLikelihoods.at(it->first) = thisAAF;
                setDAFsplit1fromLikelihoods.at(it->first) = thisAAFsplit1;
                setDAFsplit2fromLikelihoods.at(it->first) = thisAAFsplit2;
            } else if (AAint == AncestralAlleleAlt) { // Ancestral allele seems to be alt, so derived is ref
                setDAFsFromLikelihoods.at(it->first) = (1 - thisAAF);
                setDAFsplit1fromLikelihoods.at(it->first) = 1 - thisAAFsplit1;
                setDAFsplit2fromLikelihoods.at(it->first) = 1 - thisAAFsplit2;
            }
            } catch (std::out_of_range& e) { std::cerr << "The trouble was here" << it->first << std::endl; }
        }
    }
}



// Works only on biallelic markers
void GeneralSetCounts::getSetVariantCountsSimple(const std::vector<std::string>& genotypes, const std::map<size_t, string>& posToSpeciesMap) {
    // std::cerr << fields[0] << "\t" << fields[1] << std::endl;
    getBasicCounts(genotypes, posToSpeciesMap);
    
    // Now fill in the allele frequencies
    for(std::map<string,int>::iterator it = setAltCounts.begin(); it != setAltCounts.end(); ++it) {
        if (setAlleleCounts.at(it->first) > 0) {
            setAAFs[it->first] = (double)setAltCounts.at(it->first)/setAlleleCounts.at(it->first);
        }
    }
}

void GeneralSetCounts::getBasicCounts(const std::vector<std::string>& genotypes, const std::map<size_t, string>& posToSpeciesMap) {
    // Go through the genotypes - only biallelic markers are allowed
    for (std::vector<std::string>::size_type i = 0; i != genotypes.size(); i++) {
        bool speciesDefined = true;
        std::string species; try { species = posToSpeciesMap.at(i); } catch (const std::out_of_range& oor) {
            speciesDefined = false;
        }
        // The first allele in this individual
        if (genotypes[i][0] == '1') { overall++; individualsWithVariant[i]++; }
        if (genotypes[i][2] == '1') { overall++; individualsWithVariant[i]++; }
        if (speciesDefined) {
            if (genotypes[i][0] == '1') {
                setAltCounts[species]++; setAlleleCounts[species]++;
            } else if (genotypes[i][0] == '0') {
                setAlleleCounts[species]++;
            }
            // The second allele in this individual
            if (genotypes[i][2] == '1') {
                setAltCounts[species]++; setAlleleCounts[species]++;
            } else if (genotypes[i][2] == '0') {
                setAlleleCounts[species]++;
            }
        }
    }
}

void GeneralSetCountsWithSplits::getBasicCountsWithSplitsNew(const std::vector<std::string>& genotypes, const std::map<size_t, string>& posToSpeciesMap) {
    
    // Go through the genotypes - only biallelic markers are allowed
    for (std::vector<std::string>::size_type i = 0; i != genotypes.size(); i++) {
        bool speciesDefined = true;
        std::string species; try { species = posToSpeciesMap.at(i); } catch (const std::out_of_range& oor) {
            speciesDefined = false;
        }
        
        if (speciesDefined) {
            string onlyGenotypeCalls = split(genotypes[i], ':')[0];   // The string with 0/0, 0/1, 1/0, 1/1, or e.g. 0/0/1/1 for a tetraploid
            if (onlyGenotypeCalls[0] == '.') {
                continue;   // Ignore missing data
            }
            // Find ploidy
            int l = (int)onlyGenotypeCalls.length();
            int numGTs = (l/2)+1;
            setAlleleCounts[species] += numGTs;
            
            // Go through the genotypes and fill in the data structure "GeneralSetCountsWithSplits"
            for (std::vector<std::string>::size_type j = 0; j <= l; j = j+2) {
               // std::cerr << "genotypes[i][j]: " << genotypes[i][j] << std::endl;
                setGenotypes[species].push_back(genotypes[i][j] - '0');
                if (genotypes[i][j] == '1') {
                    overall++; individualsWithVariant[i]++;
                    setAltCounts[species]++;
                }
            }
            double individualAF = (double)individualsWithVariant[i]/numGTs;
            
            /* std::cerr << "onlyGenotypeCalls: " << onlyGenotypeCalls << std::endl;
            std::cerr << "individualsWithVariant[i]: " << individualsWithVariant[i] << std::endl;
            std::cerr << "numGTs: " << numGTs << std::endl;
            std::cerr << "individualAF: " << individualAF << std::endl;
            */
            setIndividualAFs[species].push_back(individualAF);
        }
    }
}

void GeneralSetCountsWithSplits::getBasicCountsFromLikelihoodsOrProbabilities(const std::vector<std::string>& genotypes, const std::map<size_t, string>& posToSpeciesMap, const int likelihoodsOrProbabilitiesTagPosition) {
    
    // Go through the genotypes - only biallelic markers are allowed
    for (std::vector<string>::size_type i = 0; i != genotypes.size(); i++) {
        bool speciesDefined = true;
        string species; try { species = posToSpeciesMap.at(i); } catch (const std::out_of_range& oor) {
            speciesDefined = false;
        }
        
        if (speciesDefined) {
            string thisLikelihoodsOrProbabilitiesString = split(genotypes[i], ':')[likelihoodsOrProbabilitiesTagPosition];
            if (thisLikelihoodsOrProbabilitiesString == ".") continue;
            else {
                setAlleleProbCounts.at(species) += 2;
                std::vector<double> thisLikelihoodsOrProbabilities = splitToDouble(thisLikelihoodsOrProbabilitiesString,',');
                std::vector<double> thisProbabilities;
                switch (likelihoodsProbabilitiesType)
                {
                    case LikelihoodsProbabilitiesPL:
                        transformFromPhred(thisLikelihoodsOrProbabilities);
                     // print_vector(thisLikelihoodsOrProbabilities, std::cerr);
                        thisProbabilities = probabilitiesFromLikelihoods(thisLikelihoodsOrProbabilities,species);
                        break;
                    case LikelihoodsProbabilitiesGL: break;
                    case LikelihoodsProbabilitiesGP:
                        thisProbabilities = thisLikelihoodsOrProbabilities;
                        break;
                }
                setIndividualExpectedGenotypes[species].push_back(getExpectedGenotype(thisProbabilities));
            }
        }
    }
}

void GeneralSetCountsWithSplits::getSplitCountsNew(const std::vector<std::string>& genotypes, const std::map<size_t, string>& posToSpeciesMap) {
    
    getBasicCountsWithSplitsNew(genotypes, posToSpeciesMap);
    
    // If at least one of the outgroup individuals has non-missing data
    // Find out what is the "ancestral allele" - i.e. the one more common in the outgroup
    try {
        if (setAlleleCounts.at("Outgroup") > 0) {
            if ((double)vector_sum(setGenotypes.at("Outgroup"))/setGenotypes.at("Outgroup").size() < 0.5) { AAint = AncestralAlleleRef; }
            else { AAint = AncestralAlleleAlt; } 
        }
    } catch (std::out_of_range& e) { AAint = -1; }
    
    // Now fill in the allele frequencies
    double totalAAF = 0; int numNonZeroCounts = 0;
    for(std::map<string,std::vector<int>>::iterator it = setGenotypes.begin(); it != setGenotypes.end(); ++it) {
        if (it->first == "") {
            std::cerr << "it->first " << it->first << "\t"; print_vector(it->second, std::cerr); std::cerr << std::endl;
        }
        std::vector<int> thisSetGenotypes = setGenotypes.at(it->first);
        std::vector<double> thisSetIndividualAFs = setIndividualAFs.at(it->first);
        
        if (thisSetGenotypes.size() > 0) {
            numNonZeroCounts++;
            int nSplit1; int nSplit2;
            double thisAAF = (double)vector_sum(thisSetGenotypes)/thisSetGenotypes.size();
           // print_vector(thisSetGenotypes, std::cerr);
           // std::cerr << "thisAAF: " << thisAAF << std::endl;
            setAAFs[it->first] = thisAAF; totalAAF += thisAAF;
            nSplit1 = setAlleleCountsSplit1.at(it->first); nSplit2 = setAlleleCountsSplit2.at(it->first);
            
            // Take care of the splits by random sampling with replacement:
            std::random_device rd;     // only used once to initialise (seed) engine
            std::mt19937 rng(rd());    // random-number engine used (Mersenne-Twister in this case)
            std::uniform_int_distribution<int> uni(0,(thisSetGenotypes.size() - 1)); // guaranteed unbiased
            std::uniform_int_distribution<int> uniAFs(0,(thisSetIndividualAFs.size() - 1)); // guaranteed unbiased
            
            std::vector<int> thisSetGenotypesSampledSplit1; std::vector<int> thisSetGenotypesSampledSplit2;
            for (int i = 0; i < thisSetGenotypes.size(); i++) {
                int random_pos_s1 = uni(rng);
                int random_pos_s2 = uni(rng);
                thisSetGenotypesSampledSplit1.push_back(thisSetGenotypes[random_pos_s1]);
                thisSetGenotypesSampledSplit2.push_back(thisSetGenotypes[random_pos_s2]);
            }
            
            std::vector<double> thisSetIndividualAFsSampledSplit1; std::vector<double> thisSetIndividualAFsSampledSplit2;
            for (int i = 0; i < thisSetIndividualAFs.size(); i++) {
                int random_pos_s1 = uniAFs(rng);
                int random_pos_s2 = uniAFs(rng);
                thisSetIndividualAFsSampledSplit1.push_back(thisSetIndividualAFs[random_pos_s1]);
                thisSetIndividualAFsSampledSplit2.push_back(thisSetIndividualAFs[random_pos_s2]);
            }
            
          //  double thisAAFsplit1 = vector_average(thisSetGenotypesSampledSplit1);
          //  double thisAAFsplit2 = vector_average(thisSetGenotypesSampledSplit2);
            double thisAAFsplit1 = vector_average(thisSetIndividualAFsSampledSplit1);
            double thisAAFsplit2 = vector_average(thisSetIndividualAFsSampledSplit2);
            setAAFsplit1[it->first] = thisAAFsplit1; setAAFsplit2[it->first] = thisAAFsplit2;
            
            // Count correction as in admixtools
            double ya = vector_sum(thisSetGenotypes); double yb = thisSetGenotypes.size() - vector_sum(thisSetGenotypes);
            double yt = (double)thisSetGenotypes.size();
            double h = ya * yb / (yt * (yt - 1.0));
            //std::cerr << "it->first: " << it->first << std::endl;
            //std::cerr << "ya: " << ya << " ; yb: " << yb << " ; yt: " << yt << std::endl;
            //std::cerr << "h: " << h << " ; h / yt: " << h / yt << std::endl;
            
            setCorrectionFactors[it->first] = h / yt;
            
           // std::cerr << "it->first " << it->first << std::endl;
            try {
            if (AAint == AncestralAlleleRef) { // Ancestral allele seems to be the ref, so derived is alt
                setDAFs[it->first] = thisAAF;
                setDAFsplit1[it->first] = thisAAFsplit1; setDAFsplit2[it->first] = thisAAFsplit2;
            } else if (AAint == AncestralAlleleAlt) { // Ancestral allele seems to be alt, so derived is ref
                setDAFs[it->first] = (1 - thisAAF);
                setDAFsplit1[it->first] = 1 - thisAAFsplit1;
                setDAFsplit2[it->first] = 1 - thisAAFsplit2;
            }
                } catch (std::out_of_range& e) { std::cerr << "The trouble was here" << it->first << std::endl; }
        }
    }
    averageAAF = totalAAF/numNonZeroCounts;
    if (AAint == AncestralAlleleRef) averageDAF = averageAAF;
    else if (AAint == AncestralAlleleAlt) averageDAF = (1 - averageAAF);
}



int GeneralSetCounts::findADtagPosition(const std::vector<std::string>& vcfLineFields) {
    
    std::vector<std::string> format = split(vcfLineFields[8], ':');
    if (format.size() == 1) return LikelihoodsProbabilitiesAbsent; // The GT tag must be present in the first place
    
    int ADTagPosition = returnFormatTagPosition(format, "AD");
    if (ADTagPosition == std::numeric_limits<int>::min()) {
        std::cerr << "Could not find the AD tag in the VCF file. This tag is requored to use the pool-seq option. Exiting ...." << std::endl;
        exit(1);
    }
    return ADTagPosition;
}


double calculateOneDs(double ABBAtotal, double BABAtotal) {
    // Get the D values
    double Dnum1 = ABBAtotal - BABAtotal;
    
    double Ddenom1 = ABBAtotal + BABAtotal;
    double D = Dnum1/Ddenom1;
    return D;
}



double* calculateThreeDs(double ABBAtotal, double BABAtotal, double BBAAtotal) {
    // Get the D values
    double Dnum1 = ABBAtotal - BABAtotal;
    double Dnum2 = ABBAtotal - BBAAtotal;
    double Dnum3 = BBAAtotal - BABAtotal;
    
    double Ddenom1 = ABBAtotal + BABAtotal;
    double Ddenom2 = ABBAtotal + BBAAtotal;
    double Ddenom3 = BBAAtotal + BABAtotal;
    static double Ds[3]; Ds[0] = Dnum1/Ddenom1; Ds[1] = Dnum2/Ddenom2; Ds[2] = Dnum3/Ddenom3;
    return Ds;
}


double stringToDouble(std::string s) {
    double d;
    std::stringstream ss(s); //turn the string into a stream
    ss >> d; //convert
    return d;
}


// Remove a single file extension from the filename
std::string stripExtension(const std::string& filename)
{
    size_t suffixPos = filename.find_last_of('.');
    if(suffixPos == std::string::npos)
        return filename; // no suffix
    else
        return filename.substr(0, suffixPos);
}


void split(const std::string &s, char delim, std::vector<std::string> &elems) {
    std::stringstream ss(s);
    std::string item;
    while (std::getline(ss, item, delim)) {
        elems.push_back(item);
    }
}

std::vector<std::string> split(const std::string &s, char delim) {
    std::vector<std::string> elems;
    split(s, delim, elems);
    return elems;
}
    
    
void splitToDouble(const std::string &s, char delim, std::vector<double> &elems) {
    std::stringstream ss(s);
    std::string item;
    while (std::getline(ss, item, delim)) {
        elems.push_back(stringToDouble(item));
    }
}

std::vector<double> splitToDouble(const std::string &s, char delim) {
    std::vector<double> elems;
    splitToDouble(s, delim, elems);
    return elems;
}

std::vector<std::string> split2(std::string s, string delim) {
    std::vector<std::string> elems;
    size_t pos = 0;
    std::string token;
    while ((pos = s.find(delim)) != std::string::npos) {
        token = s.substr(0, pos);
        elems.push_back(token);
        s.erase(0, pos + delim.length());
    }
    elems.push_back(s);
    return elems;
}


std::vector<size_t> locateSet(const std::vector<std::string>& sample_names, const std::vector<std::string>& set) {
    std::vector<size_t> setLocs;
    for (std::vector<std::string>::size_type i = 0; i != set.size(); i++) {
        std::vector<std::string>::const_iterator it = std::find(sample_names.begin(), sample_names.end(), set[i]);
        if (it == sample_names.end()) {
            std::cerr << "Did not find the sample: \"" << set[i] << "\"" << std::endl;
            print_vector(sample_names, std::cerr,',');
        } else {
            size_t loc = std::distance(sample_names.begin(), it);
            setLocs.push_back(loc);
        }
    }
    return setLocs;
}


//
std::string suffix(const std::string& seq, size_t len)
{
    assert(seq.length() >= len);
    return seq.substr(seq.length() - len);
}

// Returns true if the filename has an extension indicating it is compressed
bool isGzip(const std::string& filename)
{
    size_t suffix_length = sizeof(GZIP_EXT) - 1;
    
    // Assume files without an extension are not compressed
    if(filename.length() < suffix_length)
        return false;
    
    std::string extension = suffix(filename, suffix_length);
    return extension == GZIP_EXT;
}

// Ensure a filehandle is open
void assertFileOpen(std::ifstream& fh, const std::string& fn)
{
    if(!fh.is_open())
    {
        std::cerr << "ERROR: Could not open " << fn << " for read\n";
        exit(EXIT_FAILURE);
    }
}
// Ensure a filehandle is open
void assertFileOpen(std::ofstream& fh, const std::string& fn)
{
    if(!fh.is_open())
    {
        std::cerr << "ERROR: Could not open " << fn << " for write\n";
        exit(EXIT_FAILURE);
    }
}


void assertGZOpen(gzstreambase& gh, const std::string& fn)
{
    if(!gh.good())
    {
        std::cerr << "ERROR: Could not open " << fn << std::endl;
        exit(EXIT_FAILURE);
    }
}

void checkGenotypesExist(const std::vector<std::string>& fields, const int variantNum) {
    if (fields.size() <= NUM_NON_GENOTYPE_COLUMNS) {
        std::cerr << "ERROR: Variant " << variantNum << " in the VCF appears to be truncated."  << std::endl;
        print_vector(fields, std::cerr);
        std::cerr << "Exiting..." << std::endl; exit(1);
    }
}

// Open a file that may or may not be gzipped for reading
// The caller is responsible for freeing the handle
std::istream* createReader(const std::string& filename, std::ios_base::openmode mode)
{
    if(isGzip(filename))
    {
        igzstream* pGZ = new igzstream(filename.c_str(), mode);
        assertGZOpen(*pGZ, filename);
        return pGZ;
    }
    else
    {
        std::ifstream* pReader = new std::ifstream(filename.c_str(), mode);
        assertFileOpen(*pReader, filename);
        return pReader;
    }
}

// Open a file that may or may not be gzipped for writing
// The caller is responsible for freeing the handle
std::ostream* createWriter(const std::string& filename,
                           std::ios_base::openmode mode)
{
    if(isGzip(filename))
    {
        ogzstream* pGZ = new ogzstream(filename.c_str(), mode);
        assertGZOpen(*pGZ, filename);
        return pGZ;
    }
    else
    {
        std::ofstream* pWriter = new std::ofstream(filename.c_str(), mode);
        assertFileOpen(*pWriter, filename);
        return pWriter;
    }
}

bool file_exists(const std::string& name) {
    std::ifstream f(name.c_str());
    return f.good();
}


void assignSplits01FromAlleleFrequency(const double p, double& splitA, double& splitB) {
    double r = ((double) rand() / (RAND_MAX));
    if (r <= p) { splitA = 1; }
    double r2 = ((double) rand() / (RAND_MAX));
    if (r2 <= p) { splitB = 1; }
}