void GaussModel::setSamples()
{
LinearInterpolation::container_type & data = interpolation_.getData();
data.clear();
if (max_ == min_)
return;
data.reserve(UInt((max_ - min_) / interpolation_step_ + 1));
CoordinateType pos = min_;
for (UInt i = 0; pos < max_; ++i)
{
pos = min_ + i * interpolation_step_;
data.push_back(statistics_.normalDensity_sqrt2pi(pos));
}
// scale data so that integral over distribution equals one
// multiply sum by interpolation_step_ -> rectangular approximation of integral
IntensityType factor = scaling_ / interpolation_step_ /
std::accumulate(data.begin(), data.end(), IntensityType(0));
for (LinearInterpolation::container_type::iterator it = data.begin(); it != data.end(); ++it)
*it *= factor;
interpolation_.setScale(interpolation_step_);
interpolation_.setOffset(min_);
}
void IsotopeModel::setSamples(const EmpiricalFormula & formula)
{
typedef std::vector<DoubleReal> ContainerType;
ContainerType isotopes_exact;
isotope_distribution_ = formula.getIsotopeDistribution(max_isotope_);
isotope_distribution_.trimRight(trim_right_cutoff_);
isotope_distribution_.renormalize();
// compute the average mass (-offset)
CoordinateType isotopes_mean = 0;
Int i = 0;
for (IsotopeDistribution::iterator iter = isotope_distribution_.begin();
iter != isotope_distribution_.end(); ++iter, ++i)
{
isotopes_exact.push_back(iter->second);
isotopes_mean += iter->second * i;
}
isotopes_mean *= isotope_distance_ / charge_;
// (Need not divide by sum of probabilities, which is 1.)
///
// "stretch" the averagine isotope distribution (so we can add datapoints between isotope peaks)
///
size_t isotopes_exact_size = isotopes_exact.size();
isotopes_exact.resize(size_t((isotopes_exact_size - 1) * isotope_distance_ / interpolation_step_ + 1.6)); // round up a bit more
for (Size i = isotopes_exact_size - 1; i; --i)
{
// we don't need to move the 0-th entry
isotopes_exact[size_t(CoordinateType(i) * isotope_distance_ / interpolation_step_ / charge_ + 0.5)]
= isotopes_exact[i];
isotopes_exact[i] = 0;
}
////
// compute the Gaussian/Cauchy distribution (to be added for widening the averagine isotope distribution)
////
ContainerType peak_shape_values_y;
// fill a container with CoordinateType points (x values)
CoordinateType peak_width = 0.0;
if (param_.getValue("isotope:mode:mode") == "Gaussian")
{
// Actual width for values in the smooth table for normal distribution
peak_width = isotope_stdev_ * 4.0; // MAGIC alert, num stdev for smooth table for normal distribution
ContainerType peak_shape_values_x;
for (DoubleReal coord = -peak_width; coord <= peak_width;
coord += interpolation_step_)
{
peak_shape_values_x.push_back(coord);
}
// compute normal approximation at these CoordinateType points (y values)
Math::BasicStatistics<> normal_widening_model;
normal_widening_model.setSum(1);
normal_widening_model.setMean(0);
normal_widening_model.setVariance(isotope_stdev_ * isotope_stdev_);
normal_widening_model.normalApproximation(peak_shape_values_y, peak_shape_values_x);
}
else if (param_.getValue("isotope:mode:mode") == "Lorentzian")
{
peak_width = isotope_lorentz_fwhm_ * 8.0; // MAGIC alert: Lorentzian has infinite support, but we need to stop sampling at some point: 8*FWHM
for (DoubleReal coord = -peak_width; coord <= peak_width;
coord += interpolation_step_)
{
boost::math::cauchy_distribution<double> cauchy(0., isotope_lorentz_fwhm_ / 2.0);
double x = boost::math::pdf(cauchy, coord);
//double y = gsl_ran_cauchy_pdf(coord, isotope_lorentz_fwhm_/2.0);
peak_shape_values_y.push_back(x); //cauchy is using HWHM not FWHM
}
}
///
// fold the Gaussian/Lorentzian at each averagine peak, i.e. fill linear interpolation
///
const ContainerType & left = isotopes_exact;
const ContainerType & right = peak_shape_values_y;
ContainerType & result = interpolation_.getData();
result.clear();
SignedSize r_max = std::min(SignedSize(left.size() + right.size() - 1),
SignedSize(2 * peak_width / interpolation_step_ * max_isotope_ + 1));
result.resize(r_max, 0);
// we loop backwards because then the small products tend to come first
// (for better numerics)
for (SignedSize i = left.size() - 1; i >= 0; --i)
{
if (left[i] == 0)
continue;
for (SignedSize j = std::min(r_max - i, SignedSize(right.size())) - 1; j >= 0; --j)
{
result[i + j] += left[i] * right[j];
}
}
monoisotopic_mz_ = mean_ - isotopes_mean;
interpolation_.setMapping(interpolation_step_, peak_width / interpolation_step_, monoisotopic_mz_);
//std::cerr << "mono now: " << monoisotopic_mz_ << " mono easy: " << formula.getMonoWeight()/formula.getCharge() << "\n";
// scale data so that integral over distribution equals one
// multiply sum by interpolation_step_ -> rectangular approximation of integral
IntensityType factor = scaling_ / (interpolation_step_ * std::accumulate(result.begin(), result.end(), IntensityType(0)));
for (ContainerType::iterator iter = result.begin(); iter != result.end(); ++iter)
{
*iter *= factor;
}
}
void ExtendedIsotopeModel::setSamples()
{
// MAGIC alert, num stdev for smooth table for normal distribution
CoordinateType normal_widening_num_stdev = 4.;
// Actual width for values in the smooth table for normal distribution
CoordinateType normal_widening_width = isotope_stdev_ * normal_widening_num_stdev;
typedef std::vector<double> ContainerType;
ContainerType isotopes_exact;
CoordinateType mass = monoisotopic_mz_ * charge_;
Int C_num = Int(0.5 + mass * averagine_[C]);
Int N_num = Int(0.5 + mass * averagine_[N]);
Int O_num = Int(0.5 + mass * averagine_[O]);
Int H_num = Int(0.5 + mass * averagine_[H]);
Int S_num = Int(0.5 + mass * averagine_[S]);
String form("");
if (C_num)
form.append("C").append(String(C_num));
if (H_num)
form.append("H").append(String(H_num));
if (N_num)
form.append("N").append(String(N_num));
if (O_num)
form.append("O").append(String(O_num));
if (S_num)
form.append("S").append(String(S_num));
EmpiricalFormula formula(form);
IsotopeDistribution isotope_distribution = formula.getIsotopeDistribution(CoarseIsotopePatternGenerator(max_isotope_));
isotope_distribution.trimRight(trim_right_cutoff_);
isotope_distribution.renormalize();
// compute the average mass (-offset)
for (IsotopeDistribution::iterator iter = isotope_distribution.begin(); iter != isotope_distribution.end(); ++iter)
{
isotopes_exact.push_back(iter->getIntensity());
}
// "stretch" the averagine isotope distribution
Size isotopes_exact_size = isotopes_exact.size();
isotopes_exact.resize(Size((isotopes_exact_size - 1)
* isotope_distance_ / interpolation_step_ + 1.6)); // round up a bit more
for (Size i = isotopes_exact_size - 1; i; --i)
{
// we don't need to move the 0-th entry
isotopes_exact[Size(CoordinateType(i) *
isotope_distance_ / interpolation_step_ / charge_ + 0.5)]
= isotopes_exact[i];
isotopes_exact[i] = 0;
}
// compute the normal distribution (to be added for widening the averagine isotope distribution)
Math::BasicStatistics<> normal_widening_model;
normal_widening_model.setSum(1);
normal_widening_model.setMean(0);
normal_widening_model.setVariance(isotope_stdev_ * isotope_stdev_);
// fill a container with CoordinateType points
ContainerType normal_widening_coordinate;
for (double coord = -normal_widening_width;
coord <= normal_widening_width;
coord += interpolation_step_
)
{
normal_widening_coordinate.push_back(coord);
}
// compute normal approximation at these CoordinateType points
ContainerType normal_widening;
normal_widening_model.normalApproximation(normal_widening, normal_widening_coordinate);
// fill linear interpolation
const ContainerType & left = isotopes_exact;
const ContainerType & right = normal_widening;
ContainerType & result = interpolation_.getData();
result.clear();
Int rMax = std::min(Int(left.size() + right.size() - 1), Int(2 * normal_widening_width / interpolation_step_ * max_isotope_ + 1));
result.resize(rMax, 0);
// we loop backwards because then the small products tend to come first
// (for better numerics)
for (SignedSize i = left.size() - 1; i >= 0; --i)
{
if (left[i] == 0)
continue;
for (SignedSize j = std::min<SignedSize>(rMax - i, right.size()) - 1; j >= 0; --j)
{
result[i + j] += left[i] * right[j];
}
}
// set interpolation
interpolation_.setMapping(interpolation_step_, normal_widening_width / interpolation_step_, monoisotopic_mz_);
// scale data so that integral over distribution equals one
// multiply sum by interpolation_step_ -> rectangular approximation of integral
IntensityType factor = scaling_ / interpolation_step_ /
std::accumulate(result.begin(), result.end(), IntensityType(0));
for (ContainerType::iterator iter = result.begin(); iter != result.end(); ++iter)
{
*iter *= factor;
}
}
