//Used by matching part to prevent pass by reference. private ResultsGroup matchPassbyValue(ResultsGroup input1, CompositionHypothesisEntry comhypo) { ResultsGroup storage = new ResultsGroup(); //Pass by value, I only way I know we can do this is to pass them one by one. Yes, it is troublesome. storage.DeconRow = input1.DeconRow; storage.MostAbundant = input1.MostAbundant; storage.NumChargeStates = input1.NumChargeStates; storage.ScanDensity = input1.ScanDensity; storage.NumModiStates = input1.NumModiStates; storage.TotalVolume = input1.TotalVolume; storage.ExpectedA = input1.ExpectedA; storage.CentroidScan = input1.CentroidScan; storage.NumOfScan = input1.NumOfScan; storage.AvgSigNoise = input1.AvgSigNoise; storage.MaxScanNum = input1.MaxScanNum; storage.MinScanNum = input1.MinScanNum; storage.ScanNumList = input1.ScanNumList; storage.ChargeStateList = input1.ChargeStateList; storage.AvgSigNoiseList = input1.AvgSigNoiseList; storage.CentroidScanLR = input1.CentroidScanLR; storage.AvgAA2List = input1.AvgAA2List; storage.PredictedComposition = comhypo; storage.Match = true; return storage; }
//this "Grouping" function performs the grouping. private List<ResultsGroup> Groupings(String filename, ParametersForm.ParameterSettings modelParameters, Double Mas, List<CompositionHypothesisEntry> comhyp) { GetDeconData DeconDATA1 = new GetDeconData(); List<string> elementIDs = new List<string>(); List<string> molename = new List<string>(); for (int i = 0; i < comhyp.Count(); i++ ) { if (comhyp[i].ElementNames.Count > 0) { for (int j = 0; j < comhyp[i].ElementNames.Count(); j++) { elementIDs.Add(comhyp[i].ElementNames[j]); } for (int j = 0; j < comhyp[i].MoleculeNames.Count(); j++) { molename.Add(comhyp[i].MoleculeNames[j]); } break; } } List<DeconRow> sortedDeconData = new List<DeconRow>();; sortedDeconData = DeconDATA1.getdata(filename); //First, sort the list descendingly by its abundance. sortedDeconData = sortedDeconData.OrderByDescending(a => a.abundance).ToList(); //################Second, create a new list to store results from the first grouping.############### List<ResultsGroup> fgResults = new List<ResultsGroup>(); ResultsGroup GR2 = new ResultsGroup(); Int32 currentMaxBin = new Int32(); currentMaxBin = 1; GR2.DeconRow = sortedDeconData[0]; GR2.MostAbundant = true; GR2.NumOfScan = 1; GR2.MinScanNum = sortedDeconData[0].ScanNum; GR2.MaxScanNum = sortedDeconData[0].ScanNum; GR2.ChargeStateList = new List<int>(); GR2.ChargeStateList.Add(sortedDeconData[0].charge); GR2.AvgSigNoiseList = new List<Double>(); GR2.AvgSigNoiseList.Add(sortedDeconData[0].SignalNoiseRatio); GR2.AvgAA2List = new List<double>(); GR2.AvgAA2List.Add(sortedDeconData[0].MonoisotopicAbundance / (sortedDeconData[0].MonoisotopicPlus2Abundance + 1)); GR2.ScanNumList = new List<Int32>(); GR2.ScanNumList.Add(sortedDeconData[0].ScanNum); GR2.NumModiStates = 1; GR2.TotalVolume = sortedDeconData[0].abundance * sortedDeconData[0].fwhm; GR2.ListAbundance = new List<double>(); GR2.ListAbundance.Add(sortedDeconData[0].abundance); GR2.ListMonoMassWeight = new List<double>(); GR2.ListMonoMassWeight.Add(sortedDeconData[0].MonoisotopicMassWeight); fgResults.Add(GR2); for (int j = 1; j < sortedDeconData.Count; j++) { for (int i = 0; i < fgResults.Count; i++) { //Obtain grouping error. Note: its in ppm, so it needs to be multiplied by 0.000001. Double GroupingError = fgResults[i].DeconRow.MonoisotopicMassWeight * modelParameters.GroupingErrorEG * 0.000001; if ((sortedDeconData[j].MonoisotopicMassWeight < (fgResults[i].DeconRow.MonoisotopicMassWeight + GroupingError) && (sortedDeconData[j].MonoisotopicMassWeight > (fgResults[i].DeconRow.MonoisotopicMassWeight - GroupingError)))) { if (fgResults[i].MaxScanNum < sortedDeconData[j].ScanNum) { fgResults[i].MaxScanNum = sortedDeconData[j].ScanNum; } else if (fgResults[i].MinScanNum > sortedDeconData[j].ScanNum) { fgResults[i].MinScanNum = sortedDeconData[j].ScanNum; } fgResults[i].NumOfScan = fgResults[i].NumOfScan + 1; fgResults[i].ScanNumList.Add(sortedDeconData[j].ScanNum); fgResults[i].TotalVolume = fgResults[i].TotalVolume + sortedDeconData[j].abundance * sortedDeconData[j].fwhm; fgResults[i].ChargeStateList.Add(sortedDeconData[j].charge); fgResults[i].AvgSigNoiseList.Add(sortedDeconData[j].SignalNoiseRatio); fgResults[i].AvgAA2List.Add(sortedDeconData[j].MonoisotopicAbundance / (sortedDeconData[j].MonoisotopicPlus2Abundance + 1)); fgResults[i].ListAbundance.Add(sortedDeconData[j].abundance); fgResults[i].ListMonoMassWeight.Add(sortedDeconData[j].MonoisotopicMassWeight); break; } if (i == fgResults.Count - 1) { ResultsGroup GR = new ResultsGroup(); currentMaxBin = currentMaxBin + 1; GR.DeconRow = sortedDeconData[j]; GR.MostAbundant = true; GR.NumOfScan = 1; GR.MinScanNum = sortedDeconData[j].ScanNum; GR.MaxScanNum = sortedDeconData[j].ScanNum; GR.ChargeStateList = new List<int>(); GR.ChargeStateList.Add(sortedDeconData[j].charge); GR.AvgSigNoiseList = new List<Double>(); GR.AvgSigNoiseList.Add(sortedDeconData[j].SignalNoiseRatio); GR.AvgAA2List = new List<double>(); GR.AvgAA2List.Add(sortedDeconData[j].MonoisotopicAbundance / (sortedDeconData[j].MonoisotopicPlus2Abundance + 1)); GR.ScanNumList = new List<int>(); GR.ScanNumList.Add(sortedDeconData[j].ScanNum); GR.NumModiStates = 1; GR.TotalVolume = sortedDeconData[j].abundance * sortedDeconData[j].fwhm; GR.ListAbundance = new List<double>(); GR.ListAbundance.Add(sortedDeconData[j].abundance); GR.ListMonoMassWeight = new List<double>(); GR.ListMonoMassWeight.Add(sortedDeconData[j].MonoisotopicMassWeight); fgResults.Add(GR); } } } //Lastly calculate the Average Weighted Abundance for (int y = 0; y < fgResults.Count(); y++) { Double sumofTopPart = 0; for (int z = 0; z < fgResults[y].ListMonoMassWeight.Count(); z++) { sumofTopPart = sumofTopPart + fgResults[y].ListMonoMassWeight[z] * fgResults[y].ListAbundance[z]; } fgResults[y].DeconRow.MonoisotopicMassWeight = sumofTopPart / fgResults[y].ListAbundance.Sum(); } //######################## Here is the second grouping. ################################ fgResults = fgResults.OrderBy(o => o.DeconRow.MonoisotopicMassWeight).ToList(); if (Mas != 0) { for (int i = 0; i < fgResults.Count - 1; i++) { if (fgResults[i].MostAbundant == true) { int numModStates = 1; for (int j = i + 1; j < fgResults.Count; j++) { Double AdductTolerance = fgResults[i].DeconRow.MonoisotopicMassWeight * modelParameters.AdductToleranceEA * 0.000001; if ((fgResults[i].DeconRow.MonoisotopicMassWeight >= (fgResults[j].DeconRow.MonoisotopicMassWeight - Mas * numModStates - AdductTolerance)) && (fgResults[i].DeconRow.MonoisotopicMassWeight <= (fgResults[j].DeconRow.MonoisotopicMassWeight - Mas * numModStates + AdductTolerance))) { //obtain max and min scan number if (fgResults[i].MaxScanNum < fgResults[j].MaxScanNum) { fgResults[i].MaxScanNum = fgResults[j].MaxScanNum; } else { fgResults[i].MaxScanNum = fgResults[i].MaxScanNum; } if (fgResults[i].MinScanNum > fgResults[j].MinScanNum) { fgResults[i].MinScanNum = fgResults[j].MinScanNum; } else { fgResults[i].MinScanNum = fgResults[i].MinScanNum; } //numOfScan fgResults[i].NumOfScan = fgResults[i].NumOfScan + fgResults[j].NumOfScan; fgResults[i].ScanNumList.AddRange(fgResults[j].ScanNumList); //ChargeStateList for (int h = 0; h < fgResults[j].ChargeStateList.Count; h++) { fgResults[i].ChargeStateList.Add(fgResults[j].ChargeStateList[h]); } //avgSigNoiseList for (int h = 0; h < fgResults[j].AvgSigNoiseList.Count; h++) { fgResults[i].AvgSigNoiseList.Add(fgResults[j].AvgSigNoiseList[h]); } //avgAA2List for (int h = 0; h < fgResults[j].AvgAA2List.Count; h++) { fgResults[i].AvgAA2List.Add(fgResults[j].AvgAA2List[h]); } //numModiStates numModStates++; fgResults[i].NumModiStates = fgResults[i].NumModiStates + 1; fgResults[j].MostAbundant = false; //TotalVolume fgResults[i].TotalVolume = fgResults[i].TotalVolume + fgResults[j].TotalVolume; if (fgResults[i].DeconRow.abundance < fgResults[j].DeconRow.abundance) { fgResults[i].DeconRow = fgResults[j].DeconRow; numModStates = 1; } } else if (fgResults[i].DeconRow.MonoisotopicMassWeight < (fgResults[j].DeconRow.MonoisotopicMassWeight - (Mas + AdductTolerance * 2) * numModStates)) { //save running time. Since the list is sorted, any other mass below won't match as an adduct. break; } } } } } else { for (int i = 0; i < fgResults.Count; i++) { fgResults[i].NumModiStates = 0; } } List<ResultsGroup> sgResults = new List<ResultsGroup>(); //Implement the scan number threshold fgResults = fgResults.OrderByDescending(a => a.NumOfScan).ToList(); Int32 scanCutOff = fgResults.Count() + 1; for (int t = 0; t < fgResults.Count(); t++) { if (fgResults[t].NumOfScan < modelParameters.MinScanNumber) { scanCutOff = t; break; } } if (scanCutOff != fgResults.Count() + 1) { fgResults.RemoveRange(scanCutOff, fgResults.Count() - scanCutOff); } //############# This is the matching part. It matches the composition hypothesis with the grouped decon data.############ String[] MolNames = new String[17]; //These numOfMatches and lists are used to fit the linear regression model for Expect A: A+2. They are put here to decrease the already-int running time. Int32 numOfMatches = new Int32(); List<Double> moleWeightforA = new List<Double>(); List<Double> AARatio = new List<Double>(); //Used to obtain all available bins for centroid scan error. //Read the other lines for compTable data. fgResults = fgResults.OrderByDescending(a => a.DeconRow.MonoisotopicMassWeight).ToList(); comhyp = comhyp.OrderByDescending(b => b.MassWeight).ToList(); bool hasMatch = false; int lastMatch = 0; for (int j = 0; j < fgResults.Count; j++) { if (fgResults[j].MostAbundant == true) { lastMatch = lastMatch - 4; if (lastMatch < 0) lastMatch = 0; for (int i = lastMatch; i < comhyp.Count; i++) { Double MatchingError = comhyp[i].MassWeight * modelParameters.MatchErrorEM * 0.000001; if ((fgResults[j].DeconRow.MonoisotopicMassWeight <= (comhyp[i].MassWeight + MatchingError)) && (fgResults[j].DeconRow.MonoisotopicMassWeight >= (comhyp[i].MassWeight - MatchingError))) { ResultsGroup GR = new ResultsGroup(); GR = matchPassbyValue(fgResults[j], comhyp[i]); sgResults.Add(GR); //Stuffs for feature numOfMatches++; moleWeightforA.Add(fgResults[j].DeconRow.MonoisotopicMassWeight); AARatio.Add(fgResults[j].AvgAA2List.Average()); lastMatch = i + 1; hasMatch = true; continue; } //Since the data is sorted, there are no more matches below that row, break it. if (fgResults[j].DeconRow.MonoisotopicMassWeight > (comhyp[i].MassWeight + MatchingError)) { if (hasMatch == false) { ResultsGroup GR = new ResultsGroup(); CompositionHypothesisEntry comhypi = new CompositionHypothesisEntry(); GR = fgResults[j]; GR.Match = false; GR.PredictedComposition = comhypi; sgResults.Add(GR); lastMatch = i; break; } else { hasMatch = false; break; } } } } } //##############Last part, this is to calculate the feature data needed for logistic regression################### //Expected A and Centroid Scan Error need linear regression. The models are built here separately. //In the this model. output is the Y axis and input is X. SimpleLinearRegression AA2regression = new SimpleLinearRegression(); List<double> aainput = new List<double>(); List<double> aaoutput = new List<double>(); //Centroid Scan Error List<double> ccinput = new List<double>(); List<double> ccoutput = new List<double>(); if (numOfMatches > 3) { for (int i = 0; i < sgResults.Count; i++) { if (sgResults[i].Match == true) { if (sgResults[i].AvgAA2List.Average() != 0) { aainput.Add(sgResults[i].DeconRow.MonoisotopicMassWeight); aaoutput.Add(sgResults[i].AvgAA2List.Average()); } if (sgResults[i].DeconRow.abundance > 250) { ccoutput.Add(sgResults[i].DeconRow.ScanNum); ccinput.Add(sgResults[i].DeconRow.MonoisotopicMassWeight); } } } } else { for (int i = 0; i < sgResults.Count; i++) { if (sgResults[i].AvgAA2List.Average() != 0) { aainput.Add(sgResults[i].DeconRow.MonoisotopicMassWeight); aaoutput.Add(sgResults[i].AvgAA2List.Average()); } if (sgResults[i].DeconRow.abundance > 250) { ccoutput.Add(sgResults[i].ScanNumList.Average()); ccinput.Add(sgResults[i].DeconRow.MonoisotopicMassWeight); } } } SimpleLinearRegression CSEregression = new SimpleLinearRegression(); CSEregression.Regress(ccinput.ToArray(), ccoutput.ToArray()); AA2regression.Regress(aainput.ToArray(), aaoutput.ToArray()); //The remaining features and input them into the grouping results for (int i = 0; i < sgResults.Count; i++) { //ScanDensiy is: Number of scan divided by (max scan number – min scan number) Double ScanDensity = new Double(); Int32 MaxScanNumber = sgResults[i].MaxScanNum; Int32 MinScanNumber = sgResults[i].MinScanNum; Double NumOfScan = sgResults[i].NumOfScan; List<Int32> numChargeStatesList = sgResults[i].ChargeStateList.Distinct().ToList(); Int32 numChargeStates = numChargeStatesList.Count; Double numModiStates = sgResults[i].NumModiStates; if ((MaxScanNumber - MinScanNumber) != 0) ScanDensity = NumOfScan / (MaxScanNumber - MinScanNumber + 15); else ScanDensity = 0; //Use this scandensity for all molecules in this grouping. sgResults[i].NumChargeStates = numChargeStates; sgResults[i].ScanDensity = ScanDensity; sgResults[i].NumModiStates = numModiStates; sgResults[i].CentroidScanLR = CSEregression.Compute(sgResults[i].DeconRow.MonoisotopicMassWeight); sgResults[i].CentroidScan = Math.Abs(sgResults[i].ScanNumList.Average() - sgResults[i].CentroidScanLR); sgResults[i].ExpectedA = Math.Abs(sgResults[i].AvgAA2List.Average() - AA2regression.Compute(sgResults[i].DeconRow.MonoisotopicMassWeight)); sgResults[i].AvgSigNoise = sgResults[i].AvgSigNoiseList.Average(); } for (int i = 0; i < sgResults.Count(); i++ ) { sgResults[i].PredictedComposition.ElementNames.Clear(); sgResults[i].PredictedComposition.MoleculeNames.Clear(); if (i == sgResults.Count() - 1) { sgResults[0].PredictedComposition.ElementNames = elementIDs; sgResults[0].PredictedComposition.MoleculeNames = molename; } } return sgResults; }
//This is used to read a ResultFile public List<ResultsGroup> ReadResultsFromFile(String path) { //This code looks int, but its just repetitive code. Look for ext and you will understand. List<ResultsGroup> Ans = new List<ResultsGroup>(); List<String> molnames = new List<String>(); FileStream FS = new FileStream(path, FileMode.Open, FileAccess.Read); StreamReader read = new StreamReader(FS); String ext = Path.GetExtension(path).Replace(".", ""); if (ext == "csv") { String header = read.ReadLine(); String[] headers = header.Split(','); List<string> elementIDs = new List<string>(); //This is another older form of data if (headers[5] != "Hypothesis MW") { Boolean moreCompounds = true; int i = 17; while (moreCompounds) { if (headers[i] != "Hypothesis MW") { elementIDs.Add(headers[i]); i++; } else { moreCompounds = false; i++; } } moreCompounds = true; while (moreCompounds) { if (headers[i] != "Adduct/Replacement") { molnames.Add(headers[i]); i++; } else moreCompounds = false; } bool firstRow = true; while (read.Peek() >= 0) { //Read data String Line = read.ReadLine(); String[] Lines = Line.Split(','); //initialize new gR object ResultsGroup gR = new ResultsGroup(); DeconRow dR = new DeconRow(); CompositionHypothesisEntry cH = new CompositionHypothesisEntry(); gR.DeconRow = dR; gR.PredictedComposition = cH; //Input data if (!String.IsNullOrEmpty(Lines[0])) { if (firstRow) { gR.PredictedComposition.ElementNames = elementIDs; gR.PredictedComposition.MoleculeNames = molnames; firstRow = false; } gR.Score = Convert.ToDouble(Lines[0]); gR.DeconRow.MonoisotopicMassWeight = Convert.ToDouble(Lines[1]); gR.PredictedComposition.CompoundComposition = Lines[2]; if (String.IsNullOrEmpty(Lines[2]) || Lines[2] == "0") gR.Match = false; else gR.Match = true; gR.PredictedComposition.PepSequence = Lines[3]; gR.NumModiStates = Convert.ToDouble(Lines[5]); gR.NumChargeStates = Convert.ToInt32(Lines[6]); gR.NumOfScan = Convert.ToDouble(Lines[7]); gR.ScanDensity = Convert.ToDouble(Lines[8]); gR.ExpectedA = Convert.ToDouble(Lines[9]); gR.AvgAA2List = new List<double>(); gR.AvgAA2List.Add(Convert.ToDouble(Lines[10])); gR.TotalVolume = Convert.ToDouble(Lines[11]); gR.AvgSigNoise = Convert.ToDouble(Lines[12]); gR.CentroidScan = Convert.ToDouble(Lines[13]); gR.DeconRow.ScanNum = Convert.ToInt32(Lines[14]); gR.MaxScanNum = Convert.ToInt32(Lines[15]); gR.MinScanNum = Convert.ToInt32(Lines[16]); gR.PredictedComposition.eqCount = new Dictionary<string, int>(); int sh = 17; for (int ele = 0; ele < elementIDs.Count(); ele++ ) { gR.PredictedComposition.ElementAmount.Add(Convert.ToInt32(Lines[sh])); sh++; } gR.PredictedComposition.MassWeight = Convert.ToDouble(Lines[sh]); sh++; List<int> eqCoun = new List<int>(); for (int j = 0; j < molnames.Count(); j++) { eqCoun.Add(Convert.ToInt32(Lines[sh + j])); } gR.PredictedComposition.eqCounts = eqCoun; gR.PredictedComposition.AddRep = Lines[sh + molnames.Count()]; gR.PredictedComposition.AdductNum = Convert.ToInt32(Lines[sh + molnames.Count() + 1]); gR.PredictedComposition.PepModification = Lines[sh + molnames.Count() + 2]; gR.PredictedComposition.MissedCleavages = Convert.ToInt32(Lines[sh + molnames.Count() + 3]); gR.PredictedComposition.NumGlycosylations = Convert.ToInt32(Lines[sh + molnames.Count() + 4]); gR.PredictedComposition.StartAA = Convert.ToInt32(Lines[sh + molnames.Count() + 5]); gR.PredictedComposition.EndAA = Convert.ToInt32(Lines[sh + molnames.Count() + 6]); if (Lines.Count() > sh + molnames.Count() + 7) { gR.PredictedComposition.ProteinID = Lines[sh + molnames.Count() + 7]; } else { gR.PredictedComposition.ProteinID = "?"; } Ans.Add(gR); } } } //older data format. else if (headers[3] == "PeptideSequence") { Boolean moreCompounds = true; int i = 24; while (moreCompounds) { if (headers[i] != "Adduct/Replacement") { molnames.Add(headers[i]); i++; } else moreCompounds = false; } bool firstRow = true; while (read.Peek() >= 0) { //Read data String Line = read.ReadLine(); String[] Lines = Line.Split(','); //initialize new gR object ResultsGroup gR = new ResultsGroup(); DeconRow dR = new DeconRow(); CompositionHypothesisEntry cH = new CompositionHypothesisEntry(); gR.DeconRow = dR; gR.PredictedComposition = cH; if (firstRow) { gR.PredictedComposition.ElementNames.AddRange(new List<string> { "C", "H", "N", "O", "S", "P" }); gR.PredictedComposition.MoleculeNames = molnames; firstRow = false; } //Input data if (!String.IsNullOrEmpty(Lines[0])) { gR.Score = Convert.ToDouble(Lines[0]); gR.DeconRow.MonoisotopicMassWeight = Convert.ToDouble(Lines[1]); gR.PredictedComposition.CompoundComposition = Lines[2]; if (String.IsNullOrEmpty(Lines[2]) || Lines[2] == "0") gR.Match = false; else gR.Match = true; gR.PredictedComposition.PepSequence = Lines[3]; gR.PredictedComposition.MassWeight = Convert.ToDouble(Lines[5]); gR.NumModiStates = Convert.ToDouble(Lines[6]); gR.NumChargeStates = Convert.ToInt32(Lines[7]); gR.NumOfScan = Convert.ToDouble(Lines[8]); gR.ScanDensity = Convert.ToDouble(Lines[9]); gR.ExpectedA = Convert.ToDouble(Lines[10]); gR.AvgAA2List = new List<double>(); gR.AvgAA2List.Add(Convert.ToDouble(Lines[11])); gR.TotalVolume = Convert.ToDouble(Lines[12]); gR.AvgSigNoise = Convert.ToDouble(Lines[13]); gR.CentroidScan = Convert.ToDouble(Lines[14]); gR.DeconRow.ScanNum = Convert.ToInt32(Lines[15]); gR.MaxScanNum = Convert.ToInt32(Lines[16]); gR.MinScanNum = Convert.ToInt32(Lines[17]); gR.PredictedComposition.eqCount = new Dictionary<string, int>(); for (int k = 18; k < 24; k++) { gR.PredictedComposition.ElementAmount.Add(Convert.ToInt32(Lines[k])); } List<int> eqCoun = new List<int>(); for (int j = 0; j < molnames.Count(); j++) { eqCoun.Add(Convert.ToInt32(Lines[24 + j])); } gR.PredictedComposition.eqCounts = eqCoun; gR.PredictedComposition.AddRep = Lines[24 + molnames.Count()]; gR.PredictedComposition.AdductNum = Convert.ToInt32(Lines[24 + molnames.Count() + 1]); gR.PredictedComposition.PepModification = Lines[24 + molnames.Count() + 2]; gR.PredictedComposition.MissedCleavages = Convert.ToInt32(Lines[24 + molnames.Count() + 3]); gR.PredictedComposition.NumGlycosylations = Convert.ToInt32(Lines[24 + molnames.Count() + 4]); Ans.Add(gR); } } } //This is supporting an older format of data. Today is Sept 2013, can be deleted after 1 year. else { Boolean moreCompounds = true; int i = 23; while (moreCompounds) { if (headers[i] != "Adduct/Replacement") { molnames.Add(headers[i]); i++; } else moreCompounds = false; } bool firstRow = true; while (read.Peek() >= 0) { //Read data String Line = read.ReadLine(); String[] Lines = Line.Split(','); //initialize new gR object ResultsGroup gR = new ResultsGroup(); if (firstRow) { gR.PredictedComposition.ElementNames.AddRange(new List<string> { "C", "H", "N", "O", "S", "P" }); gR.PredictedComposition.MoleculeNames = molnames; firstRow = false; } DeconRow dR = new DeconRow(); CompositionHypothesisEntry cH = new CompositionHypothesisEntry(); gR.DeconRow = dR; gR.PredictedComposition = cH; if (!String.IsNullOrEmpty(Lines[0])) { //Input data gR.Score = Convert.ToDouble(Lines[0]); gR.DeconRow.MonoisotopicMassWeight = Convert.ToDouble(Lines[1]); gR.PredictedComposition.CompoundComposition = Lines[2].Replace(",", ";"); if (String.IsNullOrEmpty(Lines[2]) || Lines[2] == "0") gR.Match = false; else gR.Match = true; gR.PredictedComposition.MassWeight = Convert.ToDouble(Lines[4]); gR.NumModiStates = Convert.ToDouble(Lines[5]); gR.NumChargeStates = Convert.ToInt32(Lines[6]); gR.NumOfScan = Convert.ToDouble(Lines[7]); gR.ScanDensity = Convert.ToDouble(Lines[8]); gR.ExpectedA = Convert.ToDouble(Lines[9]); gR.AvgAA2List = new List<double>(); gR.AvgAA2List.Add(Convert.ToDouble(Lines[10])); gR.TotalVolume = Convert.ToDouble(Lines[11]); gR.AvgSigNoise = Convert.ToDouble(Lines[12]); gR.CentroidScan = Convert.ToDouble(Lines[13]); gR.DeconRow.ScanNum = Convert.ToInt32(Lines[14]); gR.MaxScanNum = Convert.ToInt32(Lines[15]); gR.MinScanNum = Convert.ToInt32(Lines[16]); gR.PredictedComposition.eqCount = new Dictionary<string, int>(); for (int k = 17; k < 23; k++) { gR.PredictedComposition.ElementAmount.Add(Convert.ToInt32(Lines[k])); } gR.PredictedComposition.eqCount.Add("A", Convert.ToInt32(Lines[23])); gR.PredictedComposition.eqCount.Add("B", Convert.ToInt32(Lines[24])); gR.PredictedComposition.eqCount.Add("C", Convert.ToInt32(Lines[25])); gR.PredictedComposition.eqCount.Add("D", Convert.ToInt32(Lines[26])); gR.PredictedComposition.eqCount.Add("E", Convert.ToInt32(Lines[27])); gR.PredictedComposition.eqCount.Add("F", Convert.ToInt32(Lines[28])); gR.PredictedComposition.eqCount.Add("G", Convert.ToInt32(Lines[29])); gR.PredictedComposition.eqCount.Add("H", Convert.ToInt32(Lines[30])); gR.PredictedComposition.eqCount.Add("I", Convert.ToInt32(Lines[31])); gR.PredictedComposition.eqCount.Add("J", Convert.ToInt32(Lines[32])); gR.PredictedComposition.eqCount.Add("K", Convert.ToInt32(Lines[33])); gR.PredictedComposition.eqCount.Add("L", Convert.ToInt32(Lines[34])); gR.PredictedComposition.eqCount.Add("M", Convert.ToInt32(Lines[35])); gR.PredictedComposition.eqCount.Add("N", Convert.ToInt32(Lines[36])); gR.PredictedComposition.eqCount.Add("O", Convert.ToInt32(Lines[37])); gR.PredictedComposition.eqCount.Add("P", Convert.ToInt32(Lines[38])); gR.PredictedComposition.eqCount.Add("Q", Convert.ToInt32(Lines[39])); gR.PredictedComposition.AddRep = Lines[40]; gR.PredictedComposition.AdductNum = Convert.ToInt32(Lines[41]); gR.PredictedComposition.PepSequence = Lines[42]; gR.PredictedComposition.PepModification = Lines[43]; gR.PredictedComposition.MissedCleavages = Convert.ToInt32(Lines[44]); gR.PredictedComposition.NumGlycosylations = Convert.ToInt32(Lines[45]); Ans.Add(gR); } } } } //This is gly1 data. else { String header = read.ReadLine(); String[] headers = header.Split('\t'); while (read.Peek() >= 0) { //Read data String Line = read.ReadLine(); String[] Lines = Line.Split('\t'); //initialize new gR object ResultsGroup gR = new ResultsGroup(); DeconRow dR = new DeconRow(); CompositionHypothesisEntry cH = new CompositionHypothesisEntry(); gR.DeconRow = dR; gR.PredictedComposition = cH; if (!String.IsNullOrEmpty(Lines[0])) { //Input data gR.PredictedComposition.MoleculeNames = molnames; gR.Score = Convert.ToDouble(Lines[0]); gR.DeconRow.MonoisotopicMassWeight = Convert.ToDouble(Lines[1]); gR.PredictedComposition.CompoundComposition = Lines[2].Replace(",", ";"); if (String.IsNullOrEmpty(Lines[2]) || Lines[2] == "0") { gR.Match = false; gR.PredictedComposition.MassWeight = 0; } else { gR.Match = true; gR.PredictedComposition.MassWeight = Convert.ToDouble(Lines[4]); } gR.NumModiStates = Convert.ToDouble(Lines[5]); gR.NumChargeStates = Convert.ToInt32(Lines[6]); gR.NumOfScan = Convert.ToDouble(Lines[7]); gR.ScanDensity = Convert.ToDouble(Lines[8]); gR.ExpectedA = Convert.ToDouble(Lines[9]); gR.AvgAA2List = new List<double>(); gR.AvgAA2List.Add(Convert.ToDouble(Lines[10])); gR.TotalVolume = Convert.ToDouble(Lines[11]); gR.AvgSigNoise = Convert.ToDouble(Lines[12]); gR.CentroidScan = Convert.ToDouble(Lines[13]); gR.DeconRow.ScanNum = Convert.ToInt32(Convert.ToDouble(Lines[14])); Ans.Add(gR); } } } return Ans; }
private ResultsGroup AverageResultsGroup(List<ResultsGroup> combiningRows) { List<Int32> NumChargeStates = new List<int>(); List<Double> ScanDensity= new List<Double>() ; List<Double> NumModiStates= new List<Double>() ; List<Double> TotalVolume= new List<Double>() ; List<Double> ExpectedA= new List<Double>() ; List<Double> Score= new List<Double>() ; List<Double> CentroidScan= new List<Double>() ; List<Double> NumScan= new List<Double>() ; List<Double> AvgSigNoise= new List<Double>() ; //These are for calculating the features List<Int32> MaxScanNum= new List<int>() ; List<Int32> MinScanNum= new List<int>() ; List<Int32> ScanNums= new List<int>() ; List<Int32> ChargeStateList = new List<int>(); List<Double> AvgSigNoiseList= new List<Double>() ; List<Double> CentroidScanLR= new List<Double>() ; List<Double> AvgAA2List= new List<Double>() ; List<Double> RelativeTotalVolume = new List<double>(); List<Int32> scan_num = new List<Int32>(); List<Int32> charge = new List<Int32>(); List<Int32> abundance = new List<Int32>(); List<Double> mz = new List<Double>(); List<Double> fit = new List<Double>(); List<Double> average_mw = new List<Double>(); List<Double> monoisotopic_mw = new List<Double>(); List<Double> mostabundant_mw = new List<Double>(); List<Double> fwhm = new List<Double>(); List<Double> signal_noise = new List<Double>(); List<Int32> mono_abundance = new List<Int32>(); List<Int32> mono_plus2_abundance = new List<Int32>(); List<Int32> flag = new List<Int32>(); List<Double> interference_sore = new List<Double>(); ResultsGroup FinalAns = new ResultsGroup(); FinalAns.ListOfOriginalTotalVolumes = combiningRows[0].ListOfOriginalTotalVolumes; for (int i = 0; i < combiningRows.Count(); i++) { NumChargeStates.Add(combiningRows[i].NumChargeStates); ScanDensity.Add(combiningRows[i].ScanDensity); NumModiStates.Add(combiningRows[i].NumModiStates); TotalVolume.Add(combiningRows[i].TotalVolume); ExpectedA.Add(combiningRows[i].ExpectedA); Score.Add(combiningRows[i].Score); CentroidScan.Add(combiningRows[i].CentroidScan); NumScan.Add(combiningRows[i].NumOfScan); AvgSigNoise.Add(combiningRows[i].AvgSigNoise); //These are for calculating the features MaxScanNum.Add(combiningRows[i].MaxScanNum); MinScanNum.Add(combiningRows[i].MinScanNum); CentroidScanLR.Add(combiningRows[i].CentroidScanLR); AvgAA2List.AddRange(combiningRows[i].AvgAA2List); RelativeTotalVolume.Add(combiningRows[i].RelativeTotalVolume); charge.Add(combiningRows[i].DeconRow.charge); scan_num.Add(combiningRows[i].DeconRow.ScanNum); abundance.Add(combiningRows[i].DeconRow.abundance); mz.Add(combiningRows[i].DeconRow.mz); fit.Add(combiningRows[i].DeconRow.fit); average_mw.Add(combiningRows[i].DeconRow.average_mw); monoisotopic_mw.Add(combiningRows[i].DeconRow.MonoisotopicMassWeight); mostabundant_mw.Add(combiningRows[i].DeconRow.mostabundant_mw); fwhm.Add(combiningRows[i].DeconRow.fwhm); signal_noise.Add(combiningRows[i].DeconRow.SignalNoiseRatio); mono_abundance.Add(combiningRows[i].DeconRow.MonoisotopicAbundance); mono_plus2_abundance.Add(combiningRows[i].DeconRow.MonoisotopicPlus2Abundance); flag.Add(combiningRows[i].DeconRow.flag); interference_sore.Add(combiningRows[i].DeconRow.interference_sore); for (int h = 0; h < combiningRows[i].ListOfOriginalTotalVolumes.Count(); h++) { if (combiningRows[i].ListOfOriginalTotalVolumes[h] > FinalAns.ListOfOriginalTotalVolumes[h]) { FinalAns.ListOfOriginalTotalVolumes[h] = combiningRows[i].ListOfOriginalTotalVolumes[h]; } } } FinalAns.DeconRow = new DeconRow(); FinalAns.PredictedComposition = combiningRows[0].PredictedComposition; FinalAns.NumChargeStates = Convert.ToInt32(NumChargeStates.Average()); FinalAns.ScanDensity = ScanDensity.Average(); FinalAns.NumModiStates = NumModiStates.Average(); FinalAns.TotalVolume = TotalVolume.Average(); FinalAns.ExpectedA = ExpectedA.Average(); FinalAns.Score = Score.Average(); FinalAns.CentroidScan = CentroidScan.Average(); FinalAns.NumOfScan = NumScan.Average(); FinalAns.AvgSigNoise = AvgSigNoise.Average(); FinalAns.MaxScanNum = Convert.ToInt32(MaxScanNum.Average()); FinalAns.MinScanNum = Convert.ToInt32(MinScanNum.Average()); FinalAns.CentroidScanLR = CentroidScanLR.Average(); FinalAns.TotalVolumeSD = TotalVolume.StdDev(); FinalAns.RelativeTotalVolumeSD = RelativeTotalVolume.StdDev(); FinalAns.AvgAA2List = new List<double>(); FinalAns.AvgAA2List.Add(AvgAA2List.Average()); FinalAns.RelativeTotalVolume = RelativeTotalVolume.Average(); FinalAns.DeconRow.ScanNum = Convert.ToInt32(scan_num.Average()); FinalAns.DeconRow.abundance = Convert.ToInt32(abundance.Average()); FinalAns.DeconRow.mz = mz.Average(); FinalAns.DeconRow.fit = fit.Average(); FinalAns.DeconRow.average_mw = average_mw.Average(); FinalAns.DeconRow.MonoisotopicMassWeight = monoisotopic_mw.Average(); FinalAns.DeconRow.mostabundant_mw = mostabundant_mw.Average(); FinalAns.DeconRow.fwhm = fwhm.Average(); FinalAns.DeconRow.SignalNoiseRatio = signal_noise.Average(); FinalAns.DeconRow.MonoisotopicAbundance = Convert.ToInt32(mono_abundance.Average()); FinalAns.DeconRow.MonoisotopicPlus2Abundance = Convert.ToInt32(mono_plus2_abundance.Average()); FinalAns.DeconRow.flag = Convert.ToInt32(flag.Average()); FinalAns.DeconRow.interference_sore = interference_sore.Average(); return FinalAns; }
//this Grouping function performs the grouping. private List<ResultsGroup> Groupings(String filename, ParametersForm.ParameterSettings paradata) { GetDeconData DeconDATA1 = new GetDeconData(); List<DeconRow> sortedDeconData = new List<DeconRow>(); sortedDeconData = DeconDATA1.getdata(filename); //First, sort the list descendingly by its abundance. sortedDeconData = sortedDeconData.OrderByDescending(a => a.abundance).ToList(); //################Second, create a new list to store results from the first grouping.############### List<ResultsGroup> fgResults = new List<ResultsGroup>(); ResultsGroup GR2 = new ResultsGroup(); GR2.PredictedComposition = new CompositionHypothesisEntry(); Int32 currentMaxBin = new Int32(); currentMaxBin = 1; GR2.DeconRow = sortedDeconData[0]; GR2.MostAbundant = true; GR2.NumOfScan = 1; GR2.MinScanNum = sortedDeconData[0].ScanNum; GR2.MaxScanNum = sortedDeconData[0].ScanNum; GR2.ChargeStateList = new List<int>(); GR2.ChargeStateList.Add(sortedDeconData[0].charge); GR2.AvgSigNoiseList = new List<Double>(); GR2.AvgSigNoiseList.Add(sortedDeconData[0].SignalNoiseRatio); GR2.AvgAA2List = new List<double>(); GR2.AvgAA2List.Add(sortedDeconData[0].MonoisotopicAbundance / (sortedDeconData[0].MonoisotopicPlus2Abundance + 1)); GR2.ScanNumList = new List<Int32>(); GR2.ScanNumList.Add(sortedDeconData[0].ScanNum); GR2.NumModiStates = 1; GR2.TotalVolume = sortedDeconData[0].abundance * sortedDeconData[0].fwhm; GR2.ListAbundance = new List<double>(); GR2.ListAbundance.Add(sortedDeconData[0].abundance); GR2.ListMonoMassWeight = new List<double>(); GR2.ListMonoMassWeight.Add(sortedDeconData[0].MonoisotopicMassWeight); fgResults.Add(GR2); for (int j = 1; j < sortedDeconData.Count; j++) { for (int i = 0; i < fgResults.Count; i++) { //Obtain grouping error. Note: its in ppm, so it needs to be multiplied by 0.000001. Double GroupingError = fgResults[i].DeconRow.MonoisotopicMassWeight * paradata.GroupingErrorEG * 0.000001; if ((sortedDeconData[j].MonoisotopicMassWeight < (fgResults[i].DeconRow.MonoisotopicMassWeight + GroupingError) && (sortedDeconData[j].MonoisotopicMassWeight > (fgResults[i].DeconRow.MonoisotopicMassWeight - GroupingError)))) { if (fgResults[i].MaxScanNum < sortedDeconData[j].ScanNum) { fgResults[i].MaxScanNum = sortedDeconData[j].ScanNum; } else if (fgResults[i].MinScanNum > sortedDeconData[j].ScanNum) { fgResults[i].MinScanNum = sortedDeconData[j].ScanNum; } fgResults[i].NumOfScan = fgResults[i].NumOfScan + 1; fgResults[i].ScanNumList.Add(sortedDeconData[j].ScanNum); fgResults[i].TotalVolume = fgResults[i].TotalVolume + sortedDeconData[j].abundance * sortedDeconData[j].fwhm; fgResults[i].ChargeStateList.Add(sortedDeconData[j].charge); fgResults[i].AvgSigNoiseList.Add(sortedDeconData[j].SignalNoiseRatio); fgResults[i].AvgAA2List.Add(sortedDeconData[j].MonoisotopicAbundance / (sortedDeconData[j].MonoisotopicPlus2Abundance + 1)); fgResults[i].ListAbundance.Add(sortedDeconData[j].abundance); fgResults[i].ListMonoMassWeight.Add(sortedDeconData[j].MonoisotopicMassWeight); break; } if (i == fgResults.Count - 1) { ResultsGroup GR = new ResultsGroup(); GR.PredictedComposition = new CompositionHypothesisEntry(); currentMaxBin = currentMaxBin + 1; GR.DeconRow = sortedDeconData[j]; GR.MostAbundant = true; GR.NumOfScan = 1; GR.MinScanNum = sortedDeconData[j].ScanNum; GR.MaxScanNum = sortedDeconData[j].ScanNum; GR.ChargeStateList = new List<int>(); GR.ChargeStateList.Add(sortedDeconData[j].charge); GR.AvgSigNoiseList = new List<Double>(); GR.AvgSigNoiseList.Add(sortedDeconData[j].SignalNoiseRatio); GR.AvgAA2List = new List<double>(); GR.AvgAA2List.Add(sortedDeconData[j].MonoisotopicAbundance / (sortedDeconData[j].MonoisotopicPlus2Abundance + 1)); GR.ScanNumList = new List<int>(); GR.ScanNumList.Add(sortedDeconData[j].ScanNum); GR.NumModiStates = 1; GR.TotalVolume = sortedDeconData[j].abundance * sortedDeconData[j].fwhm; GR.ListAbundance = new List<double>(); GR.ListAbundance.Add(sortedDeconData[j].abundance); GR.ListMonoMassWeight = new List<double>(); GR.ListMonoMassWeight.Add(sortedDeconData[j].MonoisotopicMassWeight); fgResults.Add(GR); } } } //Lastly calculate the Average Weighted Abundance for (int y = 0; y < fgResults.Count(); y++) { Double sumofTopPart = 0; for (int z = 0; z < fgResults[y].ListMonoMassWeight.Count(); z++) { sumofTopPart = sumofTopPart + fgResults[y].ListMonoMassWeight[z] * fgResults[y].ListAbundance[z]; } fgResults[y].DeconRow.MonoisotopicMassWeight = sumofTopPart / fgResults[y].ListAbundance.Sum(); } //######################## Here is the second grouping for NH3. ################################ fgResults = fgResults.OrderBy(o => o.DeconRow.MonoisotopicMassWeight).ToList(); for (int i = 0; i < fgResults.Count - 1; i++) { if (fgResults[i].MostAbundant == true) { int numModStates = 1; for (int j = i + 1; j < fgResults.Count; j++) { Double AdductTolerance = fgResults[i].DeconRow.MonoisotopicMassWeight * paradata.AdductToleranceEA * 0.000001; if ((fgResults[i].DeconRow.MonoisotopicMassWeight >= (fgResults[j].DeconRow.MonoisotopicMassWeight - 17.02654911 * numModStates - AdductTolerance)) && (fgResults[i].DeconRow.MonoisotopicMassWeight <= (fgResults[j].DeconRow.MonoisotopicMassWeight - 17.02654911 * numModStates + AdductTolerance))) { //obtain max and min scan number if (fgResults[i].MaxScanNum < fgResults[j].MaxScanNum) { fgResults[i].MaxScanNum = fgResults[j].MaxScanNum; } else { fgResults[i].MaxScanNum = fgResults[i].MaxScanNum; } if (fgResults[i].MinScanNum > fgResults[j].MinScanNum) { fgResults[i].MinScanNum = fgResults[j].MinScanNum; } else { fgResults[i].MinScanNum = fgResults[i].MinScanNum; } //numOfScan fgResults[i].NumOfScan = fgResults[i].NumOfScan + fgResults[j].NumOfScan; fgResults[i].ScanNumList.AddRange(fgResults[j].ScanNumList); //ChargeStateList for (int h = 0; h < fgResults[j].ChargeStateList.Count; h++) { fgResults[i].ChargeStateList.Add(fgResults[j].ChargeStateList[h]); } //avgSigNoiseList for (int h = 0; h < fgResults[j].AvgSigNoiseList.Count; h++) { fgResults[i].AvgSigNoiseList.Add(fgResults[j].AvgSigNoiseList[h]); } //avgAA2List for (int h = 0; h < fgResults[j].AvgAA2List.Count; h++) { fgResults[i].AvgAA2List.Add(fgResults[j].AvgAA2List[h]); } //numModiStates numModStates++; fgResults[i].NumModiStates = fgResults[i].NumModiStates + 1; fgResults[j].MostAbundant = false; //TotalVolume fgResults[i].TotalVolume = fgResults[i].TotalVolume + fgResults[j].TotalVolume; if (fgResults[i].DeconRow.abundance < fgResults[j].DeconRow.abundance) { fgResults[i].DeconRow = fgResults[j].DeconRow; numModStates = 1; } } else if (fgResults[i].DeconRow.MonoisotopicMassWeight < (fgResults[j].DeconRow.MonoisotopicMassWeight - (17.02654911 + AdductTolerance * 2) * numModStates)) { //save running time. Since the list is sorted, any other mass below won't match as an adduct. break; } } } } //Implement the scan number threshold fgResults = fgResults.OrderByDescending(a => a.NumOfScan).ToList(); Int32 scanCutOff = fgResults.Count() + 1; for (int t = 0; t < fgResults.Count(); t++) { if (fgResults[t].NumOfScan < paradata.MinScanNumber) { scanCutOff = t; break; } } if (scanCutOff != fgResults.Count() + 1) { fgResults.RemoveRange(scanCutOff, fgResults.Count() - scanCutOff); } for (int i = 0; i < fgResults.Count(); i++) { fgResults[i].Match = false; } //##############Last part, this is to calculate the feature data needed for logistic regression################### //Expected A and Centroid Scan Error need linear regression. The models are built here separately. //In the this model. output is the Y axis and input is X. SimpleLinearRegression AA2regression = new SimpleLinearRegression(); List<double> aainput = new List<double>(); List<double> aaoutput = new List<double>(); //Centroid Scan Error List<double> ccinput = new List<double>(); List<double> ccoutput = new List<double>(); for (int i = 0; i < fgResults.Count; i++) { if (fgResults[i].AvgAA2List.Average() != 0) { aainput.Add(fgResults[i].DeconRow.MonoisotopicMassWeight); aaoutput.Add(fgResults[i].AvgAA2List.Average()); } if (fgResults[i].DeconRow.abundance > 250) { ccoutput.Add(fgResults[i].ScanNumList.Average()); ccinput.Add(fgResults[i].DeconRow.MonoisotopicMassWeight); } } SimpleLinearRegression CSEregression = new SimpleLinearRegression(); CSEregression.Regress(ccinput.ToArray(), ccoutput.ToArray()); AA2regression.Regress(aainput.ToArray(), aaoutput.ToArray()); //The remaining features and input them into the grouping results for (int i = 0; i < fgResults.Count; i++) { //ScanDensiy is: Number of scan divided by (max scan number – min scan number) Double ScanDensity = new Double(); Int32 MaxScanNumber = fgResults[i].MaxScanNum; Int32 MinScanNumber = fgResults[i].MinScanNum; Double NumOfScan = fgResults[i].NumOfScan; List<Int32> numChargeStatesList = fgResults[i].ChargeStateList.Distinct().ToList(); Int32 numChargeStates = numChargeStatesList.Count; Double numModiStates = fgResults[i].NumModiStates; if ((MaxScanNumber - MinScanNumber) != 0) ScanDensity = NumOfScan / (MaxScanNumber - MinScanNumber + 15); else ScanDensity = 0; //Use this scandensity for all molecules in this grouping. fgResults[i].NumChargeStates = numChargeStates; fgResults[i].ScanDensity = ScanDensity; fgResults[i].NumModiStates = numModiStates; fgResults[i].CentroidScanLR = CSEregression.Compute(fgResults[i].DeconRow.MonoisotopicMassWeight); fgResults[i].CentroidScan = Math.Abs(fgResults[i].ScanNumList.Average() - fgResults[i].CentroidScanLR); fgResults[i].ExpectedA = Math.Abs(fgResults[i].AvgAA2List.Average() - AA2regression.Compute(fgResults[i].DeconRow.MonoisotopicMassWeight)); fgResults[i].AvgSigNoise = fgResults[i].AvgSigNoiseList.Average(); } return fgResults; }