ExitCodes main_(int, const char**)
{
//-------------------------------------------------------------
// parameter handling
//-------------------------------------------------------------
StringList in = getStringList_("in");
String edta = getStringOption_("pos");
String out = getStringOption_("out");
String out_sep = getStringOption_("out_separator");
String out_TIC_debug = getStringOption_("auto_rt:out_debug_TIC");
StringList in_header = getStringList_("in_header");
// number of out_debug_TIC files and input files must be identical
/*if (out_TIC_debug.size() > 0 && in.size() != out_TIC_debug.size())
{
LOG_FATAL_ERROR << "Error: number of input file 'in' and auto_rt:out_debug_TIC files must be identical!" << std::endl;
return ILLEGAL_PARAMETERS;
}*/
// number of header files and input files must be identical
if (in_header.size() > 0 && in.size() != in_header.size())
{
LOG_FATAL_ERROR << "Error: number of input file 'in' and 'in_header' files must be identical!" << std::endl;
return ILLEGAL_PARAMETERS;
}
if (!getFlag_("auto_rt:enabled") && !out_TIC_debug.empty())
{
LOG_FATAL_ERROR << "Error: TIC output file requested, but auto_rt is not enabled! Either do not request the file or switch on 'auto_rt:enabled'." << std::endl;
return ILLEGAL_PARAMETERS;
}
double rttol = getDoubleOption_("rt_tol");
double mztol = getDoubleOption_("mz_tol");
Size rt_collect = getIntOption_("rt_collect");
//-------------------------------------------------------------
// loading input
//-------------------------------------------------------------
MzMLFile mzml_file;
mzml_file.setLogType(log_type_);
MSExperiment<Peak1D> exp, exp_pp;
EDTAFile ed;
ConsensusMap cm;
ed.load(edta, cm);
StringList tf_single_header0, tf_single_header1, tf_single_header2; // header content, for each column
std::vector<String> vec_single; // one line for each compound, multiple columns per experiment
vec_single.resize(cm.size());
for (Size fi = 0; fi < in.size(); ++fi)
{
// load raw data
mzml_file.load(in[fi], exp);
exp.sortSpectra(true);
if (exp.empty())
{
LOG_WARN << "The given file does not contain any conventional peak data, but might"
" contain chromatograms. This tool currently cannot handle them, sorry." << std::endl;
return INCOMPATIBLE_INPUT_DATA;
}
// try to detect RT peaks (only for the first input file -- all others should align!)
// cm.size() might change in here...
if (getFlag_("auto_rt:enabled") && fi == 0)
{
ConsensusMap cm_local = cm; // we might have different RT peaks for each map if 'auto_rt' is enabled
cm.clear(false); // reset global list (about to be filled)
// compute TIC
MSChromatogram<> tic = exp.getTIC();
MSSpectrum<> tics, tic_gf, tics_pp, tics_sn;
for (Size ic = 0; ic < tic.size(); ++ic)
{ // rewrite Chromatogram to MSSpectrum (GaussFilter requires it)
Peak1D peak;
peak.setMZ(tic[ic].getRT());
peak.setIntensity(tic[ic].getIntensity());
tics.push_back(peak);
}
// smooth (no PP_CWT here due to efficiency reasons -- large FWHM take longer!)
double fwhm = getDoubleOption_("auto_rt:FHWM");
GaussFilter gf;
Param p = gf.getParameters();
p.setValue("gaussian_width", fwhm * 2); // wider than FWHM, just to be sure we have a fully smoothed peak. Merging two peaks is unlikely
p.setValue("use_ppm_tolerance", "false");
gf.setParameters(p);
tic_gf = tics;
gf.filter(tic_gf);
// pick peaks
PeakPickerHiRes pp;
p = pp.getParameters();
p.setValue("signal_to_noise", getDoubleOption_("auto_rt:SNThreshold"));
pp.setParameters(p);
pp.pick(tic_gf, tics_pp);
if (tics_pp.size())
{
LOG_INFO << "Found " << tics_pp.size() << " auto-rt peaks at: ";
for (Size ipp = 0; ipp != tics_pp.size(); ++ipp) LOG_INFO << " " << tics_pp[ipp].getMZ();
}
else
{
LOG_INFO << "Found no auto-rt peaks. Change threshold parameters!";
}
LOG_INFO << std::endl;
if (!out_TIC_debug.empty()) // if debug file was given
{ // store intermediate steps for debug
MSExperiment<> out_debug;
out_debug.addChromatogram(toChromatogram(tics));
out_debug.addChromatogram(toChromatogram(tic_gf));
SignalToNoiseEstimatorMedian<MSSpectrum<> > snt;
snt.init(tics);
for (Size is = 0; is < tics.size(); ++is)
{
Peak1D peak;
peak.setMZ(tic[is].getMZ());
peak.setIntensity(snt.getSignalToNoise(tics[is]));
tics_sn.push_back(peak);
}
out_debug.addChromatogram(toChromatogram(tics_sn));
out_debug.addChromatogram(toChromatogram(tics_pp));
// get rid of "native-id" missing warning
for (Size id = 0; id < out_debug.size(); ++id) out_debug[id].setNativeID(String("spectrum=") + id);
mzml_file.store(out_TIC_debug, out_debug);
LOG_DEBUG << "Storing debug AUTO-RT: " << out_TIC_debug << std::endl;
}
// add target EICs: for each m/z with no/negative RT, add all combinations of that m/z with auto-RTs
// duplicate m/z entries will be ignored!
// all other lines with positive RT values are copied unaffected
//do not allow doubles
std::set<double> mz_doubles;
for (ConsensusMap::Iterator cit = cm_local.begin(); cit != cm_local.end(); ++cit)
{
if (cit->getRT() < 0)
{
if (mz_doubles.find(cit->getMZ()) == mz_doubles.end())
{
mz_doubles.insert(cit->getMZ());
}
else
{
LOG_INFO << "Found duplicate m/z entry (" << cit->getMZ() << ") for auto-rt. Skipping ..." << std::endl;
continue;
}
ConsensusMap cm_RT_multiplex;
for (MSSpectrum<>::ConstIterator itp = tics_pp.begin(); itp != tics_pp.end(); ++itp)
{
ConsensusFeature f = *cit;
f.setRT(itp->getMZ());
cm.push_back(f);
}
}
else
{ // default feature with no auto-rt
LOG_INFO << "copying feature with RT " << cit->getRT() << std::endl;
cm.push_back(*cit);
}
}
// resize, since we have more positions now
vec_single.resize(cm.size());
}
// search for each EIC and add up
Int not_found(0);
Map<Size, double> quant;
String description;
if (fi < in_header.size())
{
HeaderInfo info(in_header[fi]);
description = info.header_description;
}
if (fi == 0)
{ // two additional columns for first file (theoretical RT and m/z)
tf_single_header0 << "" << "";
tf_single_header1 << "" << "";
tf_single_header2 << "RT" << "mz";
}
// 5 entries for each input file
tf_single_header0 << File::basename(in[fi]) << "" << "" << "" << "";
tf_single_header1 << description << "" << "" << "" << "";
tf_single_header2 << "RTobs" << "dRT" << "mzobs" << "dppm" << "intensity";
for (Size i = 0; i < cm.size(); ++i)
{
//std::cerr << "Rt" << cm[i].getRT() << " mz: " << cm[i].getMZ() << " R " << cm[i].getMetaValue("rank") << "\n";
double mz_da = mztol * cm[i].getMZ() / 1e6; // mz tolerance in Dalton
MSExperiment<>::ConstAreaIterator it = exp.areaBeginConst(cm[i].getRT() - rttol / 2,
cm[i].getRT() + rttol / 2,
cm[i].getMZ() - mz_da,
cm[i].getMZ() + mz_da);
Peak2D max_peak;
max_peak.setIntensity(0);
max_peak.setRT(cm[i].getRT());
max_peak.setMZ(cm[i].getMZ());
for (; it != exp.areaEndConst(); ++it)
{
if (max_peak.getIntensity() < it->getIntensity())
{
max_peak.setIntensity(it->getIntensity());
max_peak.setRT(it.getRT());
max_peak.setMZ(it->getMZ());
}
}
double ppm = 0; // observed m/z offset
if (max_peak.getIntensity() == 0)
{
++not_found;
}
else
{
// take median for m/z found
std::vector<double> mz;
MSExperiment<>::Iterator itm = exp.RTBegin(max_peak.getRT());
SignedSize low = std::min<SignedSize>(std::distance(exp.begin(), itm), rt_collect);
SignedSize high = std::min<SignedSize>(std::distance(itm, exp.end()) - 1, rt_collect);
MSExperiment<>::AreaIterator itt = exp.areaBegin((itm - low)->getRT() - 0.01, (itm + high)->getRT() + 0.01, cm[i].getMZ() - mz_da, cm[i].getMZ() + mz_da);
for (; itt != exp.areaEnd(); ++itt)
{
mz.push_back(itt->getMZ());
//std::cerr << "ppm: " << itt.getRT() << " " << itt->getMZ() << " " << itt->getIntensity() << std::endl;
}
if ((SignedSize)mz.size() > (low + high + 1)) LOG_WARN << "Compound " << i << " has overlapping peaks [" << mz.size() << "/" << low + high + 1 << "]" << std::endl;
if (!mz.empty())
{
double avg_mz = std::accumulate(mz.begin(), mz.end(), 0.0) / double(mz.size());
//std::cerr << "avg: " << avg_mz << "\n";
ppm = (avg_mz - cm[i].getMZ()) / cm[i].getMZ() * 1e6;
}
}
// appending the second column set requires separator
String append_sep = (fi == 0 ? "" : out_sep);
vec_single[i] += append_sep; // new line
if (fi == 0)
{
vec_single[i] += String(cm[i].getRT()) + out_sep +
String(cm[i].getMZ()) + out_sep;
}
vec_single[i] += String(max_peak.getRT()) + out_sep +
String(max_peak.getRT() - cm[i].getRT()) + out_sep +
String(max_peak.getMZ()) + out_sep +
String(ppm) + out_sep +
String(max_peak.getIntensity());
}
if (not_found) LOG_INFO << "Missing peaks for " << not_found << " compounds in file '" << in[fi] << "'.\n";
}
//-------------------------------------------------------------
// create header
//-------------------------------------------------------------
vec_single.insert(vec_single.begin(), ListUtils::concatenate(tf_single_header2, out_sep));
vec_single.insert(vec_single.begin(), ListUtils::concatenate(tf_single_header1, out_sep));
vec_single.insert(vec_single.begin(), ListUtils::concatenate(tf_single_header0, out_sep));
//-------------------------------------------------------------
// writing output
//-------------------------------------------------------------
TextFile tf;
for (std::vector<String>::iterator v_it = vec_single.begin(); v_it != vec_single.end(); ++v_it)
{
tf.addLine(*v_it);
}
tf.store(out);
return EXECUTION_OK;
}
void PeakPickerHiRes::pick(const MSSpectrum& input, MSSpectrum& output, std::vector<PeakBoundary>& boundaries, bool check_spacings) const
{
// copy meta data of the input spectrum
output.clear(true);
output.SpectrumSettings::operator=(input);
output.MetaInfoInterface::operator=(input);
output.setRT(input.getRT());
output.setMSLevel(input.getMSLevel());
output.setName(input.getName());
output.setType(SpectrumSettings::CENTROID);
if (report_FWHM_)
{
output.getFloatDataArrays().resize(1);
output.getFloatDataArrays()[0].setName( report_FWHM_as_ppm_ ? "FWHM_ppm" : "FWHM");
}
// don't pick a spectrum with less than 5 data points
if (input.size() < 5) return;
// if both spacing constraints are disabled, don't check spacings at all:
if ((spacing_difference_ == std::numeric_limits<double>::infinity()) &&
(spacing_difference_gap_ == std::numeric_limits<double>::infinity()))
{
check_spacings = false;
}
// signal-to-noise estimation
SignalToNoiseEstimatorMedian<MSSpectrum > snt;
snt.setParameters(param_.copy("SignalToNoise:", true));
if (signal_to_noise_ > 0.0)
{
snt.init(input);
}
// find local maxima in profile data
for (Size i = 2; i < input.size() - 2; ++i)
{
double central_peak_mz = input[i].getMZ(), central_peak_int = input[i].getIntensity();
double left_neighbor_mz = input[i - 1].getMZ(), left_neighbor_int = input[i - 1].getIntensity();
double right_neighbor_mz = input[i + 1].getMZ(), right_neighbor_int = input[i + 1].getIntensity();
// do not interpolate when the left or right support is a zero-data-point
if (std::fabs(left_neighbor_int) < std::numeric_limits<double>::epsilon()) continue;
if (std::fabs(right_neighbor_int) < std::numeric_limits<double>::epsilon()) continue;
// MZ spacing sanity checks
double left_to_central = 0.0, central_to_right = 0.0, min_spacing = 0.0;
if (check_spacings)
{
left_to_central = central_peak_mz - left_neighbor_mz;
central_to_right = right_neighbor_mz - central_peak_mz;
min_spacing = (left_to_central < central_to_right) ? left_to_central : central_to_right;
}
double act_snt = 0.0, act_snt_l1 = 0.0, act_snt_r1 = 0.0;
if (signal_to_noise_ > 0.0)
{
act_snt = snt.getSignalToNoise(input[i]);
act_snt_l1 = snt.getSignalToNoise(input[i - 1]);
act_snt_r1 = snt.getSignalToNoise(input[i + 1]);
}
// look for peak cores meeting MZ and intensity/SNT criteria
if ((central_peak_int > left_neighbor_int) &&
(central_peak_int > right_neighbor_int) &&
(act_snt >= signal_to_noise_) &&
(act_snt_l1 >= signal_to_noise_) &&
(act_snt_r1 >= signal_to_noise_) &&
(!check_spacings ||
((left_to_central < spacing_difference_ * min_spacing) &&
(central_to_right < spacing_difference_ * min_spacing))))
{
// special case: if a peak core is surrounded by more intense
// satellite peaks (indicates oscillation rather than
// real peaks) -> remove
double act_snt_l2 = 0.0, act_snt_r2 = 0.0;
if (signal_to_noise_ > 0.0)
{
act_snt_l2 = snt.getSignalToNoise(input[i - 2]);
act_snt_r2 = snt.getSignalToNoise(input[i + 2]);
}
// checking signal-to-noise?
if ((i > 1) &&
(i + 2 < input.size()) &&
(left_neighbor_int < input[i - 2].getIntensity()) &&
(right_neighbor_int < input[i + 2].getIntensity()) &&
(act_snt_l2 >= signal_to_noise_) &&
(act_snt_r2 >= signal_to_noise_) &&
(!check_spacings ||
((left_neighbor_mz - input[i - 2].getMZ() < spacing_difference_ * min_spacing) &&
(input[i + 2].getMZ() - right_neighbor_mz < spacing_difference_ * min_spacing))))
{
++i;
continue;
}
std::map<double, double> peak_raw_data;
peak_raw_data[central_peak_mz] = central_peak_int;
peak_raw_data[left_neighbor_mz] = left_neighbor_int;
peak_raw_data[right_neighbor_mz] = right_neighbor_int;
// peak core found, now extend it
// to the left
Size k = 2;
bool previous_zero_left(false); // no need to extend peak if previous intensity was zero
Size missing_left(0);
Size left_boundary(i - 1); // index of the left boundary for the spline interpolation
while ((k <= i) && // prevent underflow
(i - k + 1 > 0) &&
!previous_zero_left &&
(missing_left <= missing_) &&
(input[i - k].getIntensity() <= peak_raw_data.begin()->second) &&
(!check_spacings ||
(peak_raw_data.begin()->first - input[i - k].getMZ() < spacing_difference_gap_ * min_spacing)))
{
double act_snt_lk = 0.0;
if (signal_to_noise_ > 0.0)
{
act_snt_lk = snt.getSignalToNoise(input[i - k]);
}
if ((act_snt_lk >= signal_to_noise_) &&
(!check_spacings ||
(peak_raw_data.begin()->first - input[i - k].getMZ() < spacing_difference_ * min_spacing)))
{
peak_raw_data[input[i - k].getMZ()] = input[i - k].getIntensity();
}
else
{
++missing_left;
if (missing_left <= missing_)
{
peak_raw_data[input[i - k].getMZ()] = input[i - k].getIntensity();
}
}
previous_zero_left = (input[i - k].getIntensity() == 0);
left_boundary = i - k;
++k;
}
// to the right
k = 2;
bool previous_zero_right(false); // no need to extend peak if previous intensity was zero
Size missing_right(0);
Size right_boundary(i+1); // index of the right boundary for the spline interpolation
while ((i + k < input.size()) &&
!previous_zero_right &&
(missing_right <= missing_) &&
(input[i + k].getIntensity() <= peak_raw_data.rbegin()->second) &&
(!check_spacings ||
(input[i + k].getMZ() - peak_raw_data.rbegin()->first < spacing_difference_gap_ * min_spacing)))
{
double act_snt_rk = 0.0;
if (signal_to_noise_ > 0.0)
{
act_snt_rk = snt.getSignalToNoise(input[i + k]);
}
if ((act_snt_rk >= signal_to_noise_) &&
(!check_spacings ||
(input[i + k].getMZ() - peak_raw_data.rbegin()->first < spacing_difference_ * min_spacing)))
{
peak_raw_data[input[i + k].getMZ()] = input[i + k].getIntensity();
}
else
{
++missing_right;
if (missing_right <= missing_)
{
peak_raw_data[input[i + k].getMZ()] = input[i + k].getIntensity();
}
}
previous_zero_right = (input[i + k].getIntensity() == 0);
right_boundary = i + k;
++k;
}
// skip if the minimal number of 3 points for fitting is not reached
if (peak_raw_data.size() < 3) continue;
CubicSpline2d peak_spline (peak_raw_data);
// calculate maximum by evaluating the spline's 1st derivative
// (bisection method)
double max_peak_mz = central_peak_mz;
double max_peak_int = central_peak_int;
double threshold = 1e-6;
OpenMS::Math::spline_bisection(peak_spline, left_neighbor_mz, right_neighbor_mz, max_peak_mz, max_peak_int, threshold);
//
// compute FWHM
//
if (report_FWHM_)
{
double fwhm_int = max_peak_int / 2.0;
threshold = 0.01 * fwhm_int;
double mz_mid, int_mid;
// left:
double mz_left = peak_raw_data.begin()->first;
double mz_center = max_peak_mz;
if (peak_spline.eval(mz_left) > fwhm_int)
{ // the spline ends before half max is reached -- take the leftmost point (probably an underestimation)
mz_mid = mz_left;
}
else
{
do
{
mz_mid = mz_left / 2 + mz_center / 2;
int_mid = peak_spline.eval(mz_mid);
if (int_mid < fwhm_int)
{
mz_left = mz_mid;
}
else
{
mz_center = mz_mid;
}
} while (fabs(int_mid - fwhm_int) > threshold);
}
const double fwhm_left_mz = mz_mid;
// right ...
double mz_right = peak_raw_data.rbegin()->first;
mz_center = max_peak_mz;
if (peak_spline.eval(mz_right) > fwhm_int)
{ // the spline ends before half max is reached -- take the rightmost point (probably an underestimation)
mz_mid = mz_right;
}
else
{
do
{
mz_mid = mz_right / 2 + mz_center / 2;
int_mid = peak_spline.eval(mz_mid);
if (int_mid < fwhm_int)
{
mz_right = mz_mid;
}
else
{
mz_center = mz_mid;
}
} while (fabs(int_mid - fwhm_int) > threshold);
}
const double fwhm_right_mz = mz_mid;
const double fwhm_absolute = fwhm_right_mz - fwhm_left_mz;
output.getFloatDataArrays()[0].push_back( report_FWHM_as_ppm_ ? fwhm_absolute / max_peak_mz * 1e6 : fwhm_absolute);
} // FWHM
// save picked peak into output spectrum
Peak1D peak;
PeakBoundary peak_boundary;
peak.setMZ(max_peak_mz);
peak.setIntensity(max_peak_int);
peak_boundary.mz_min = input[left_boundary].getMZ();
peak_boundary.mz_max = input[right_boundary].getMZ();
output.push_back(peak);
boundaries.push_back(peak_boundary);
// jump over profile data points that have been considered already
i = i + k - 1;
}
}
return;
}