/**
* put the pointers into an array sorted by mass
*
* @param Peptol the precursor mass tolerance
* @param Zdep should the tolerance be charge dependent?
* @return maximum m/z value
*/
int CMSPeakSet::SortPeaks(int Peptol, int Zdep, int Numisotopes, bool Pepppm, int ChargeSign)
{
int iCharges;
CMSPeak* Peaks;
TPeakSet::iterator iPeakSet;
int CalcMass; // the calculated mass
TMassPeak *temp;
int ptol; // charge corrected mass tolerance
int MaxMZ(0);
MassIntervals.Clear();
// first sort
for(iPeakSet = GetPeaks().begin();
iPeakSet != GetPeaks().end();
iPeakSet++ ) {
Peaks = *iPeakSet;
// skip empty spectra
if(Peaks->GetError() == eMSHitError_notenuffpeaks) continue;
// loop thru possible charges
for(iCharges = 0; iCharges < Peaks->GetNumCharges(); iCharges++) {
// correction for incorrect charge determination.
// see 12/13/02 notebook, pg. 135
if(Pepppm) {
ptol = (Zdep * (Peaks->GetCharges()[iCharges] - 1) + 1);
ptol *= Peptol;
ptol *= MSSCALE2DBL(Peaks->GetPrecursormz()) / 1000000.0;
}
else
ptol = (Zdep * (Peaks->GetCharges()[iCharges] - 1) + 1) * Peptol;
int iIsotopes;
for(iIsotopes = 0; iIsotopes <= Numisotopes; ++iIsotopes) {
temp = new TMassPeak;
CalcMass = Peaks->GetPrecursormz() * Peaks->GetCharges()[iCharges] -
MSSCALE2INT(Peaks->GetCharges()[iCharges]*kProton) * ChargeSign;
temp->ExpMass = CalcMass;
CalcMass = CalcMass - MSSCALE2INT(iIsotopes*kNeutron); // correction for c-13
temp->Mass = CalcMass;
temp->Peptol = ptol;
temp->Charge = Peaks->GetCharges()[iCharges];
temp->Peak = Peaks;
// save the TMassPeak info
const CRange<ncbi::TSignedSeqPos> myrange(temp->Mass - temp->Peptol, temp->Mass + temp->Peptol);
const ncbi::CConstRef<ncbi::CObject> myobject(static_cast <CObject *> (temp));
MassIntervals.Insert(myrange,
myobject);
// keep track of maximum m/z
if(temp->Mass + temp->Peptol > MaxMZ)
MaxMZ = temp->Mass + temp->Peptol;
}
}
}
return MaxMZ;
}
void Clonality::AddClonalPenalty(std::vector<float> const& signalInFlow, std::vector<int> const& keyLen, int const fnum, std::vector<int>& flowCount, std::vector<float>& penalty)
{
// adds the penalty for every bead in this flow
// characterize the flow using signal across beads
std::vector<double> peaks;
GetPeaks(signalInFlow, peaks);
scprint( this, "adjusted_peaks=%d;\n", (int)peaks.size());
// debug_peaks.assign(peaks.begin(), peaks.end());
size_t npeaks = peaks.size();
if (npeaks < 2) { // this flow needs to have at least 2 peaks to be used
return;
}
// if signal is too low a bead will rejected
float too_low = peaks[0] -(peaks[1] - peaks[0])*.8;
scprint( this,"too_low = peaks[0] = %f - (peaks[1]=%f -peaks[0])*.8 = %f\n", peaks[0], peaks[1], too_low);
// beads with signals near the 5th peak are 4-mers, ignore higher
float missing = (npeaks < 5) ? 5 - npeaks : 0;
float too_high = peaks[npeaks-1] + (missing + .5)*(peaks[1]-peaks[0]);
scprint( this,"too_high = peaks[%d] = %f + (missing +.5 = %f) * (peaks[1] = %f - peaks[0] = %f) = %f\n", (int)npeaks-1, peaks[npeaks-1], (missing+.5), peaks[1], peaks[0], too_high);
// given the flow characteristics, add the flow penalty to the overall penalty
// for each bead. If a penalty is applied, also increment flowCount for the bead
for (int ibd=0; ibd < (int)signalInFlow.size(); ibd++){
// key flows don't get penalized
if (fnum < keyLen[ibd])
continue;
// maximum penalty already applied
#if __cplusplus >= 201103L
if ( std::isinf(penalty[ibd]) )
#else
if ( isinf(penalty[ibd]) )
#endif
continue;
// reject any bead where signal is not present by using maximum penalty
if ( isnan( signalInFlow[ibd] )) {
penalty[ibd] = numeric_limits<float>::infinity();
continue;
}
// reject any beads with signals that are too low using maximum penalty
if (signalInFlow[ibd] < too_low) {
// fprintf(stdout, "signalInFlow[%d]=%f < %f\n", (int)ibd, signalInFlow[ibd], too_low);
penalty[ibd] = numeric_limits<float>::infinity();
continue;
}
// find the nearest peak and calculate the penalty
bool found = false;
for (size_t i=0; i < npeaks-1; i++) {
if ( (signalInFlow[ibd] >= peaks[i]*1.5 -peaks[i+1]*.5) &&
(signalInFlow[ibd] < (peaks[i]+peaks[i+1])*.5) ) {
float w = 1.0f/(i+1); // weight some peaks more heavily?
penalty[ibd] += (signalInFlow[ibd]-peaks[i])*(signalInFlow[ibd]-peaks[i])*w;
flowCount[ibd]++;
found = true;
break;
}
}
if (found)
continue;
if ( (signalInFlow[ibd] >= (peaks[npeaks-2]+peaks[npeaks-1])*.5 ) &&
(signalInFlow[ibd] < (peaks[npeaks-1]*1.5 - peaks[npeaks-1]*.5) ) )
{
float w = 1.0f/npeaks;
penalty[ibd] += (signalInFlow[ibd]-peaks[npeaks-1])*(signalInFlow[ibd]-peaks[npeaks-1])*w;
flowCount[ibd]++;
continue;
}
// do not increase the penalty if signal is higher than known peaks
// but not too high. Do not penalize the bead in this flow
if ( (signalInFlow[ibd] >= (peaks[npeaks-1]*1.5 - peaks[npeaks-2]*.5) ) &&
(signalInFlow[ibd] < too_high))
{
continue;
}
// reject any beads with signals that are too high
if (signalInFlow[ibd] >= too_high){
// fprintf(stdout, "signalInFlow[%d]=%f > %f\n", (int)ibd, signalInFlow[ibd], too_high);
penalty[ibd] = numeric_limits<float>::infinity();
continue;
}
}
}
