/// <summary>
/// Extract isomer ratios from a given spectrum (transformed into capacity vector)
/// </summary>
/// <param name="ratiosToFit"></param>
/// <param name="nbProductsToKeep"></param>
/// <param name="capacity"></param>
/// <param name="tolerance"></param>
/// <param name="PrecursorIntensityInCTrap"></param>
/// <param name="PrecursorIntensity"></param>
/// <param name="underFlow"></param>
/// <param name="percentError"></param>
/// <param name="ConSole"></param>
/// <returns></returns>
public static Dictionary <CharacterizedPrecursor, SolvedResult> SolveFromSpectrum(IEnumerable <CharacterizedPrecursor> ratiosToFit, int nbProductsToKeep,
IEnumerable <MsMsPeak> capacity, MassTolerance tolerance,
double PrecursorIntensityInCTrap, double PrecursorIntensity,
out double underFlow, out double percentError, IConSol ConSole, string fileOut = null)
{
bool keepGoing = true;
Dictionary <double, double> mixedSpectrum = new Dictionary <double, double>();
List <Dictionary <double, double> > unitSpectrum = new List <Dictionary <double, double> >();
foreach (CharacterizedPrecursor isomer in ratiosToFit)
{
foreach (double key in isomer.NormalizedFragments[nbProductsToKeep].Keys)
{
if (!mixedSpectrum.ContainsKey(key))
{
double cumulIntensity = 0.0;
foreach (MsMsPeak peak in capacity)
{
if (Math.Abs(Utilities.Numerics.CalculateMassError(peak.MZ, key, tolerance.Units)) <= tolerance.Value)
{
cumulIntensity += peak.Intensity;
}
}
mixedSpectrum.Add(key, cumulIntensity);// / PrecursorIntensityInCTrap);
}
}
if (isomer.NormalizedFragments.ContainsKey(nbProductsToKeep))
{
if (isomer.FragmentNormalizor.ContainsKey(nbProductsToKeep))
{
Dictionary <double, double> dic = new Dictionary <double, double>();
foreach (double key in isomer.NormalizedFragments[nbProductsToKeep].Keys)
{
dic.Add(key, isomer.NormalizedFragments[nbProductsToKeep][key] *
isomer.FragmentNormalizor[nbProductsToKeep].InterpolateIntensity(PrecursorIntensityInCTrap));
}
unitSpectrum.Add(dic);
}
else
{
unitSpectrum.Add(isomer.NormalizedFragments[nbProductsToKeep]);
}
}
else
{
keepGoing = false;
}
}
vsCSVWriter writerFrag = null;
if (!string.IsNullOrEmpty(fileOut))
{
writerFrag = new vsCSVWriter(fileOut);
string line = "Fragments:";
foreach (double key in mixedSpectrum.Keys)
{
line += "," + key;
}
writerFrag.AddLine(line);
line = "Mixed:";
foreach (double val in mixedSpectrum.Values)
{
line += "," + val;
}
writerFrag.AddLine(line);
}
//This nbProduct seems relevant, try to use isomer to get ratios for this spectrum
if (keepGoing)
{
List <double> solution = new List <double>();
double stepSize = PrecursorIntensityInCTrap / 1000.0;
if (stepSize < 1)
{
stepSize = 1;
}
double tmpUnderflow = 0;
Utilities.Methods.GradientDescent.SolveMaxFlowStyle(unitSpectrum, mixedSpectrum, out solution, out tmpUnderflow, ConSole, stepSize);
//Utilities.Methods.GradientDescent.SolveFromGradientDescent(unitSpectrum, mixedSpectrum, PrecursorIntensityInCTrap, out solution, out tmpUnderflow, ConSole);
double sumOfIntensities = 0;
foreach (double val in mixedSpectrum.Values)
{
sumOfIntensities += val;
}
underFlow = tmpUnderflow;
List <SolvedResult> result = GetResultList(solution, underFlow, sumOfIntensities);
Dictionary <CharacterizedPrecursor, SolvedResult> resultPerSample = new Dictionary <CharacterizedPrecursor, SolvedResult>();
int i = 0;
foreach (CharacterizedPrecursor key in ratiosToFit)
{
resultPerSample.Add(key, result[i]);
i++;
}
if (writerFrag != null)
{
foreach (CharacterizedPrecursor cPrec in ratiosToFit)
{
string line = cPrec.Peptide.Sequence;
foreach (double key in mixedSpectrum.Keys)
{
line += "," + cPrec.NormalizedFragments[nbProductsToKeep][key] * resultPerSample[cPrec].NbFitTimes;
}
writerFrag.AddLine(line);
}
writerFrag.WriteToFile();
}
percentError = (underFlow / sumOfIntensities);
return(resultPerSample);
}
else
{
percentError = 1.0;
underFlow = 0;
return(new Dictionary <CharacterizedPrecursor, SolvedResult>());
}
}
/// <summary>
/// Compares individual items with their respective items from control file.
/// </summary>
/// <param name="items">Items to compare.</param>
/// <param name="sampleToControlMaxFold">Maximum sample to control ratio.</param>
/// <param name="controlToSampleMaxFold">Maximum control to sample ratio.</param>
/// <param name="massTolerance">Mass tolerance used for peak consolidation.</param>
/// <param name="inputFiles">Input files information sorted in the final view order.</param>
protected override Dictionary <UnknownFeatureIonInstanceItem, Tuple <InControlStatus, double?> > CompareComponentItems(
List <UnknownFeatureIonInstanceItem> items,
double sampleToControlMaxFold,
double controlToSampleMaxFold,
MassTolerance massTolerance,
List <InputFileInfo> inputFiles)
{
// init results dict
var results = new Dictionary <UnknownFeatureIonInstanceItem, Tuple <InControlStatus, double?> >();
// make input files map
var inputFilesMap = inputFiles.ToDictionary(key => key.FileID);
// get all control items
var controlItems = items.Where(w => inputFilesMap[w.FileID].IsReference).ToList();
// compare all items
foreach (var item in items)
{
// init status and ratio
InControlStatus status;
double? ratio = null;
// get input file info
var inputFile = inputFilesMap[item.FileID];
// control item itself but NOT found
if (inputFile.IsReference && item.Area.Equals(0))
{
status = InControlStatus.NotInControlSelf;
results.Add(item, Tuple.Create(status, ratio));
continue;
}
// control item itself
if (inputFile.IsReference)
{
status = InControlStatus.InControlSelf;
results.Add(item, Tuple.Create(status, ratio));
continue;
}
// no control file assigned
if (inputFile.HasReference == false)
{
status = InControlStatus.NoControlAssigned;
results.Add(item, Tuple.Create(status, ratio));
continue;
}
// get corresponding control item
double maxDelta = massTolerance.GetToleranceInU(item.MolecularWeight);
var controlItem = controlItems.FirstOrDefault(w => w.FileID == inputFile.ReferenceFileID && Math.Abs(w.MolecularWeight - item.MolecularWeight) <= maxDelta);
// not found in control
if (controlItem == null || controlItem.Area.Equals(0))
{
status = InControlStatus.NotInControl;
results.Add(item, Tuple.Create(status, ratio));
continue;
}
// calc sample to control area ratio
ratio = item.Area / controlItem.Area;
// get in-control status
bool inControl = (sampleToControlMaxFold.Equals(0) || ratio <= sampleToControlMaxFold) &&
(controlToSampleMaxFold.Equals(0) || 1.0 / ratio <= controlToSampleMaxFold);
status = (inControl) ? InControlStatus.InControl : InControlStatus.Outside;
// store result
results.Add(item, Tuple.Create(status, ratio));
}
return(results);
}
public FastOverlapDemultiplexer(MsDataFileImpl demuxProcessedFile)
{
_dataFile = demuxProcessedFile;
_dataFile.EnableCaching(200);
_fragmentTolerance = new MassTolerance(10);
}
private static double ComputeMaxFlow(List <List <ProductMatch> > spikedMatches,
List <MsMsPeak> mixedSpectrum, MassTolerance tolerance,
ref List <List <double> > optimalSolutions,
ref double percentError,
ref List <long> average, IConSol ConSole)
{
//Create dictionnary of usefull peaks
Dictionary <float, double> mixedFragDic = new Dictionary <float, double>();
foreach (List <ProductMatch> fragmentRatio in spikedMatches)
{
foreach (ProductMatch match in fragmentRatio)
{
if (!mixedFragDic.ContainsKey((float)match.theoMz))
{
float closest = -1;
foreach (float key in mixedFragDic.Keys)
{
if (Math.Abs(Utilities.Numerics.CalculateMassError(match.theoMz, key, tolerance.Units)) <= tolerance.Value)
{
closest = key;
}
}
if (closest > 0)
{
//ConSole.WriteLine("Potential problem with selected fragment masses ");
match.theoMz = closest;
}
else
{
mixedFragDic.Add((float)match.theoMz, 0);
}
}
}
}
//Fill dictionnary with Intensities
List <float> keys = new List <float>(mixedFragDic.Keys);
foreach (MsMsPeak peak in mixedSpectrum)
{
foreach (float mz in keys)
{
if (Math.Abs(Utilities.Numerics.CalculateMassError(peak.MZ, mz, tolerance.Units)) <= tolerance.Value)
{
mixedFragDic[mz] += peak.Intensity;
}
}
}
List <long> localFlows = new List <long>();
foreach (List <ProductMatch> fragmentRatio in spikedMatches)
{
localFlows.Add(FindLocalMaximumFlow(fragmentRatio, mixedFragDic));
}
Dictionary <float, double> virtualSpectrum = BuildVirtualSpectrum(spikedMatches, localFlows, mixedFragDic);
double overError = MaxFlowHelper.ComputeOverflow(virtualSpectrum, mixedFragDic);
double underError = MaxFlowHelper.ComputeUnderflow(virtualSpectrum, mixedFragDic);
double[] bestIndexes = new double[spikedMatches.Count];
int iterSize = 1;
double bestOverallError = double.MaxValue;
List <long> bestLocalFlows = new List <long>();
Random rnd = new Random();
while (overError >= 1 && iterSize < 10000)//anything less than 1 is an acceptable solution
{
for (int index = 0; index < bestIndexes.Length; index++)
{
bestIndexes[index] = -1;
}
for (int i = 0; i < spikedMatches.Count; i++)
{
if (localFlows[i] > 0)
{
localFlows[i] -= iterSize;
virtualSpectrum = BuildVirtualSpectrum(spikedMatches, localFlows, mixedFragDic);
double tmpErrorMinus = MaxFlowHelper.ComputeUnderflow(virtualSpectrum, mixedFragDic);
double tmpErrorPlus = MaxFlowHelper.ComputeOverflow(virtualSpectrum, mixedFragDic);
double tmpFlowRate = Math.Abs(overError - tmpErrorPlus);
double underDiff = Math.Abs(underError - tmpErrorMinus);
if (underDiff >= 1)
{
tmpFlowRate /= underDiff;
}
bestIndexes[i] = tmpFlowRate;
localFlows[i] += iterSize;
}
}
//Pick pseudo randomly best index
double worstFlowRate = 0.0;
for (int index = 0; index < bestIndexes.Length; index++)
{
if (bestIndexes[index] > worstFlowRate)
{
worstFlowRate = bestIndexes[index];
}
}
if (worstFlowRate > 0)
{
int nbMatching = 0;
for (int index = 0; index < bestIndexes.Length; index++)
{
if (bestIndexes[index] >= worstFlowRate)
{
nbMatching++;
}
}
int iterChoice = rnd.Next(0, nbMatching - 1);
int iterNb = 0;
for (int index = 0; index < bestIndexes.Length; index++)
{
if (bestIndexes[index] >= worstFlowRate)
{
if (iterChoice == iterNb)
{
localFlows[index] -= iterSize;
}
iterNb++;
}
}
iterSize = 1;
}
else
{
iterSize++;
}
virtualSpectrum = BuildVirtualSpectrum(spikedMatches, localFlows, mixedFragDic);
overError = MaxFlowHelper.ComputeOverflow(virtualSpectrum, mixedFragDic);
underError = MaxFlowHelper.ComputeUnderflow(virtualSpectrum, mixedFragDic);
if (overError + underError < bestOverallError)
{
bestLocalFlows = new List <long>(localFlows);
bestOverallError = overError + underError;
}
}//End of while overflow > 1
optimalSolutions.Clear();
List <double> newList = new List <double>();
foreach (long localFlow in localFlows)
{
newList.Add(localFlow);
}
optimalSolutions.Add(newList);
newList = new List <double>();
foreach (long localFlow in bestLocalFlows)
{
newList.Add(localFlow);
}
optimalSolutions.Add(newList);
//Compute average
if (bestOverallError < double.MaxValue)
{
average.Clear();
for (int i = 0; i < optimalSolutions[0].Count; i++)
{
double sum = 0.0;
foreach (List <double> solution in optimalSolutions)
{
sum += solution[i];
}
double avg = sum / (double)optimalSolutions.Count;
average.Add((long)avg);
}
}
//Compute expected error in percentage
double sumOfIntensities = 0;
foreach (double val in mixedFragDic.Values)
{
sumOfIntensities += val;
}
percentError = underError / sumOfIntensities;
virtualSpectrum = BuildVirtualSpectrum(spikedMatches, localFlows, mixedFragDic);
return(MaxFlowHelper.ComputeUnderflow(virtualSpectrum, mixedFragDic));
}
public static Dictionary <CharacterizedPrecursor, SolvedResult> SolveFromSpectrumBKP(IEnumerable <CharacterizedPrecursor> ratiosToFit, int nbProductsToKeep,
long precision, IEnumerable <MsMsPeak> capacity, MassTolerance tolerance,
int returnType,//0 for max flow, 1 for best flow, 2 for average
double PrecursorIntensityInCTrap,
ref double overFlow, ref double underFlow, ref double errorInPercent, IConSol ConSole)
{
List <List <double> > solutions = new List <List <double> >();
List <long> average = new List <long>();
List <MsMsPeak> expandedCapacity = new List <MsMsPeak>();
double sumOfProducts = 0;
foreach (MsMsPeak peak in capacity)
{
double intensityNormed = peak.Intensity / PrecursorIntensityInCTrap;
expandedCapacity.Add(new MsMsPeak(peak.MZ, intensityNormed * precision, peak.Charge));
sumOfProducts += peak.Intensity;
}
List <List <ProductMatch> > tmpRatiosToFit = new List <List <ProductMatch> >();
//foreach (List<ProductMatch> list in ratiosToFit.Values)
foreach (CharacterizedPrecursor prec in ratiosToFit)
{
List <ProductMatch> pms = new List <ProductMatch>();
foreach (ProductMatch pm in prec.Fragments[nbProductsToKeep])
{
ProductMatch newPm = new ProductMatch(pm);
newPm.obsIntensity = newPm.normalizedIntensity;// *PrecursorIntensityInCTrap;
pms.Add(newPm);
}
tmpRatiosToFit.Add(pms);
}
double error = ComputeMaxFlow(tmpRatiosToFit, expandedCapacity, tolerance, ref solutions, ref errorInPercent, ref average, ConSole);
double sumOfIntensities = 0;
foreach (MsMsPeak peak in expandedCapacity)
{
sumOfIntensities += peak.Intensity;
}
overFlow = 0;
underFlow = error;
List <SolvedResult> result = null;
switch (returnType)
{
case 0:
result = GetResultList(solutions[0], precision, underFlow, sumOfIntensities);
break;
case 1:
result = GetResultList(solutions[1], precision, underFlow, sumOfIntensities);
break;
case 2:
List <double> tmpAverage = new List <double>();
foreach (double val in average)
{
tmpAverage.Add(val);
}
result = GetResultList(tmpAverage, precision, underFlow, sumOfIntensities);
break;
}
Dictionary <CharacterizedPrecursor, SolvedResult> resultPerSample = new Dictionary <CharacterizedPrecursor, SolvedResult>();
int i = 0;
foreach (CharacterizedPrecursor key in ratiosToFit)
{
resultPerSample.Add(key, result[i]);
i++;
}
return(resultPerSample);
}
public static Dictionary <PeptideSpectrumMatch, SolvedResult> SolveFromFragmentScore(int nbProductsToKeep,
double precision, IEnumerable <MsMsPeak> capacity, MassTolerance tolerance,
double PrecursorIntensityInCTrap,
Query query,
out double underFlow, out double percentError, IConSol ConSole)
{
List <Dictionary <double, double> > unitSpectrum = new List <Dictionary <double, double> >();
foreach (PeptideSpectrumMatch psm in query.psms)
{
double[] arrayFrag = psm.ComputeAACoverage();
Dictionary <double, double> individualSpectrum = new Dictionary <double, double>();
for (int i = 0; i < arrayFrag.Length; i++)
{
individualSpectrum.Add(i, arrayFrag[i]);
}
unitSpectrum.Add(individualSpectrum);
}
Dictionary <double, double> mixedSpectrum = new Dictionary <double, double>();
string seq = query.psms[0].Peptide.BaseSequence;
for (int i = 0; i < seq.Length; i++)
{
mixedSpectrum.Add(i, precision);
}
List <double> solution = new List <double>();
double tmpUnderflow = 0;
Utilities.Methods.GradientDescent.SolveMaxFlowStyle(unitSpectrum, mixedSpectrum, out solution, out tmpUnderflow, ConSole, 1);
double sumOfIntensities = 0;
foreach (double val in mixedSpectrum.Values)
{
sumOfIntensities += val;
}
underFlow = tmpUnderflow;
List <SolvedResult> result = GetResultList(solution, precision, underFlow, sumOfIntensities);
Dictionary <PeptideSpectrumMatch, SolvedResult> resultPerSample = new Dictionary <PeptideSpectrumMatch, SolvedResult>();
int k = 0;
foreach (PeptideSpectrumMatch key in query.psms)
{
resultPerSample.Add(key, result[k]);
k++;
}
percentError = (underFlow / sumOfIntensities);
return(resultPerSample);
}
public static Dictionary <PeptideSpectrumMatch, SolvedResult> SolveFromFragmentScoreTheoMZ(int nbProductsToKeep,
double precision, IEnumerable <MsMsPeak> capacity, MassTolerance tolerance,
double PrecursorIntensityInCTrap,
Query query,
out double underFlow, out double percentError, IConSol ConSole)
{
List <Dictionary <double, double> > unitSpectrum = new List <Dictionary <double, double> >();
foreach (PeptideSpectrumMatch psm in query.psms)
{
double[] arrayFrag = psm.ComputeAACoverage();
Dictionary <double, double> individualSpectrum = new Dictionary <double, double>();
foreach (ProductMatch match in psm.AllProductMatches)
{
if (!individualSpectrum.ContainsKey(match.theoMz))
{
individualSpectrum.Add(match.theoMz, 0);
}
double addedFactor = individualSpectrum[match.theoMz];
if (match.Fragment == null)
{
addedFactor += arrayFrag[match.fragmentPos];
}
else if (match.Fragment.IsReverse)
{
for (int i = match.fragmentPos - 1; i < arrayFrag.Length; i++)
{
addedFactor += arrayFrag[i];
}
}
else
{
for (int i = match.fragmentPos - 1; i >= 0; i--)
{
addedFactor += arrayFrag[i];
}
}
individualSpectrum[match.theoMz] += addedFactor;
}
double sum = 0.0;
foreach (double val in individualSpectrum.Values)
{
sum += val;
}
foreach (double key in new List <double>(individualSpectrum.Keys))
{
individualSpectrum[key] /= sum;
}
unitSpectrum.Add(individualSpectrum);
}
Dictionary <double, double> mixedSpectrum = new Dictionary <double, double>();
foreach (Dictionary <double, double> dic in unitSpectrum)
{
foreach (double key in dic.Keys)
{
if (!mixedSpectrum.ContainsKey(key))
{
mixedSpectrum.Add(key, precision);
}
}
}
double nbKeys = (double)mixedSpectrum.Count;
foreach (double key in new List <double>(mixedSpectrum.Keys))
{
mixedSpectrum[key] /= nbKeys;
}
//string seq = query.psms[0].Peptide.BaseSequence;
//for (int i = 0; i < seq.Length; i++)
// mixedSpectrum.Add(i, precision);
List <double> solution = new List <double>();
double tmpUnderflow = 0;
Utilities.Methods.GradientDescent.SolveMaxFlowStyle(unitSpectrum, mixedSpectrum, out solution, out tmpUnderflow, ConSole, 1);
double sumOfIntensities = 0;
foreach (double val in mixedSpectrum.Values)
{
sumOfIntensities += val;
}
underFlow = tmpUnderflow;
List <SolvedResult> result = GetResultList(solution, precision, underFlow, sumOfIntensities);
Dictionary <PeptideSpectrumMatch, SolvedResult> resultPerSample = new Dictionary <PeptideSpectrumMatch, SolvedResult>();
int k = 0;
foreach (PeptideSpectrumMatch key in query.psms)
{
resultPerSample.Add(key, result[k]);
k++;
}
percentError = (underFlow / sumOfIntensities);
return(resultPerSample);
}
public static Dictionary <CharacterizedPrecursor, SolvedResult> SolveFromSpectrum(IEnumerable <CharacterizedPrecursor> ratiosToFit, int nbProductsToKeep,
double precision, IEnumerable <MsMsPeak> capacity, MassTolerance tolerance,
double PrecursorIntensityInCTrap,
out double underFlow, out double percentError, IConSol ConSole)
{
bool keepGoing = true;
Dictionary <double, double> mixedSpectrum = new Dictionary <double, double>();
List <Dictionary <double, double> > unitSpectrum = new List <Dictionary <double, double> >();
foreach (CharacterizedPrecursor isomer in ratiosToFit)
{
foreach (double key in isomer.NormalizedFragments[nbProductsToKeep].Keys)
{
if (!mixedSpectrum.ContainsKey(key))
{
double cumulIntensity = 0.0;
foreach (MsMsPeak peak in capacity)
{
if (Math.Abs(Proteomics.Utilities.Numerics.CalculateMassError(peak.MZ, key, tolerance.Units)) <= tolerance.Value)
{
cumulIntensity += peak.Intensity;
}
}
mixedSpectrum.Add(key, cumulIntensity * precision / PrecursorIntensityInCTrap);
}
}
if (isomer.NormalizedFragments.ContainsKey(nbProductsToKeep))
{
unitSpectrum.Add(isomer.NormalizedFragments[nbProductsToKeep]);
}
else
{
keepGoing = false;
}
}
if (keepGoing)
{
List <double> solution = new List <double>();
double tmpUnderflow = 0;
Proteomics.Utilities.Methods.GradientDescent.SolveMaxFlowStyle(unitSpectrum, mixedSpectrum, out solution, out tmpUnderflow, ConSole);
double sumOfIntensities = 0;
foreach (double val in mixedSpectrum.Values)
{
sumOfIntensities += val;
}
underFlow = tmpUnderflow;
List <SolvedResult> result = GetResultList(solution, precision, underFlow, sumOfIntensities);
Dictionary <CharacterizedPrecursor, SolvedResult> resultPerSample = new Dictionary <CharacterizedPrecursor, SolvedResult>();
int i = 0;
foreach (CharacterizedPrecursor key in ratiosToFit)
{
resultPerSample.Add(key, result[i]);
i++;
}
percentError = (underFlow / sumOfIntensities);
return(resultPerSample);
}
else
{
percentError = 1.0;
underFlow = 0;
return(new Dictionary <CharacterizedPrecursor, SolvedResult>());
}
}
