void gcta::blup_snp_dosage()
{
check_autosome();
if(_mu.empty()) calcu_mu();
int i=0, j=0, k=0, col_num=_varcmp_Py.cols();
// Substract each element by 2p
for(i=0; i<_keep.size(); i++){
for(j=0; j<_include.size(); j++) _geno_dose[i][_include[j]]-=_mu[_include[j]];
}
// Calcuate A matrix
cout<<"Calculating the BLUP solution to SNP effects using imputed dosage scores ... "<<endl;
vector<double> var_SNP(_include.size()); // variance of each SNP, 2pq
eigenMatrix b_SNP=eigenMatrix::Zero(_include.size(), col_num); // variance of each SNP, 2pq
for(j=0; j<_include.size(); j++){
for(i=0; i<_keep.size(); i++) var_SNP[j]+=_geno_dose[i][_include[j]]*_geno_dose[i][_include[j]];
var_SNP[j]/=(double)(_keep.size()-1);
if(fabs(var_SNP[j])<1.0e-50) var_SNP[j]=0.0;
else var_SNP[j]=1.0/var_SNP[j];
}
for(k=0; k<_include.size(); k++){
for(i=0; i<_keep.size(); i++){
for(j=0; j<col_num; j++) b_SNP(k,j)+=_geno_dose[i][_include[k]]*_varcmp_Py(i,j);
}
for(j=0; j<col_num; j++) b_SNP(k,j)=b_SNP(k,j)*var_SNP[k]/(double)_include.size();
}
output_blup_snp(b_SNP);
}
void gcta::read_indi_blup(string blup_indi_file)
{
vector< vector<string> > g_buf;
ifstream i_indi_blup(blup_indi_file.c_str());
if(!i_indi_blup) throw("Error: can not open the file ["+blup_indi_file+"] to read.");
string str_buf, id_buf;
vector<string> id, vs_buf;
int i=0, j=0, k=0, col_num=0;
while(i_indi_blup){
i_indi_blup>>str_buf;
if(i_indi_blup.eof()) break;
id_buf=str_buf+":";
i_indi_blup>>str_buf;
id_buf+=str_buf;
getline(i_indi_blup, str_buf);
col_num=StrFunc::split_string(str_buf, vs_buf);
if(col_num<1) continue;
id.push_back(id_buf);
g_buf.push_back(vs_buf);
}
i_indi_blup.close();
update_id_map_kp(id, _id_map, _keep);
map<string, int> uni_id_map;
map<string, int>::iterator iter;
for(i=0; i<_keep.size(); i++) uni_id_map.insert(pair<string,int>(_fid[_keep[i]]+":"+_pid[_keep[i]], i));
_varcmp_Py.setZero(_keep.size(), col_num/2);
for(i=0; i<id.size(); i++){
iter=uni_id_map.find(id[i]);
if(iter==uni_id_map.end()) continue;
for(j=0, k=0; j<col_num; j+=2, k++) _varcmp_Py(iter->second,k)=atof(g_buf[i][j].c_str());
}
cout<<"BLUP solution to the total genetic effects for "<<_keep.size()<<" individuals have been read from ["+blup_indi_file+"]."<<endl;
}
// blue estimate of SNP effect
void gcta::blup_snp_geno()
{
check_autosome();
if(_mu.empty()) calcu_mu();
int i=0, j=0, k=0, col_num=_varcmp_Py.cols();
double x=0.0, fcount=0.0;
// Calcuate A matrix
cout<<"Calculating the BLUP solution to SNP effects ..."<<endl;
vector<double> var_SNP(_include.size());
eigenMatrix b_SNP=eigenMatrix::Zero(_include.size(), col_num); // variance of each SNP, 2pq
for(j=0; j<_include.size(); j++){
var_SNP[j]=_mu[_include[j]]*(1.0-0.5*_mu[_include[j]]);
if(fabs(var_SNP[j])<1.0e-50) var_SNP[j]=0.0;
else var_SNP[j]=1.0/var_SNP[j];
}
for(k=0; k<_include.size(); k++){
fcount=0.0;
for(i=0; i<_keep.size(); i++){
if(!_snp_1[_include[k]][i] || _snp_2[_include[k]][i]){
if(_allele1[_include[k]]==_ref_A[_include[k]]) x=_snp_1[_include[k]][i]+_snp_2[_include[k]][i];
else x=2.0-(_snp_1[_include[k]][i]+_snp_2[_include[k]][i]);
x=(x-_mu[_include[k]]);
for(j=0; j<col_num; j++) b_SNP(k,j)+=x*_varcmp_Py(i,j);
fcount+=1.0;
}
}
for(j=0; j<col_num; j++) b_SNP(k,j)=(b_SNP(k,j)*var_SNP[k]/fcount)*((double)_keep.size()/(double)_include.size());
}
output_blup_snp(b_SNP);
}
