private static IEnumerable <LCIMSMSFeature> SplitByScanLCGap(IEnumerable <LCIMSMSFeature> lcImsMsFeatureEnumerable)
{
var lcimsmsFeatureList = new List <LCIMSMSFeature>();
int gapSizeMax = Settings.LCGapSizeMax;
foreach (var lcimsmsFeature in lcImsMsFeatureEnumerable)
{
var sortByScanLCQuery = from imsMsFeature in lcimsmsFeature.imsMsFeatureList
orderby imsMsFeature.ScanLC
select imsMsFeature;
LCIMSMSFeature newLCImsMsFeature = null;
var scanLCReference = -99;
foreach (var imsMsFeature in sortByScanLCQuery)
{
if (imsMsFeature.ScanLC - scanLCReference - 1 <= gapSizeMax && newLCImsMsFeature != null)
{
newLCImsMsFeature.AddImsMsFeature(imsMsFeature);
}
else
{
newLCImsMsFeature = new LCIMSMSFeature(imsMsFeature.Charge);
newLCImsMsFeature.AddImsMsFeature(imsMsFeature);
lcimsmsFeatureList.Add(newLCImsMsFeature);
}
scanLCReference = imsMsFeature.ScanLC;
}
}
return(lcimsmsFeatureList);
}
public static void MergeLCIMSMSFeatures(LCIMSMSFeature dominantFeature, LCIMSMSFeature recessiveFeature)
{
var referenceMass = dominantFeature.imsMsFeatureList[0].MSFeatureList[0].MassMonoisotopic;
var massToChange = recessiveFeature.imsMsFeatureList[0].MSFeatureList[0].MassMonoisotopic;
var massChange = (int)Math.Round(referenceMass - massToChange);
var scanLcToImsMsFeatureMap = dominantFeature.imsMsFeatureList.ToDictionary(dominantImsMsFeature => dominantImsMsFeature.ScanLC);
foreach (var recessiveImsMsFeature in recessiveFeature.imsMsFeatureList)
{
// First correct the Mass of the recessive imsMsFeature
foreach (var msFeature in recessiveImsMsFeature.MSFeatureList)
{
msFeature.MassMonoisotopic += massChange;
// TODO: Keep this??
//msFeature.Mz = (msFeature.MassMonoisotopic / msFeature.Charge) + (float)1.00727849;
}
if (scanLcToImsMsFeatureMap.TryGetValue(recessiveImsMsFeature.ScanLC, out var dominantImsMsFeature))
{
MergeImsMsFeatures(dominantImsMsFeature, recessiveImsMsFeature);
}
else
{
dominantFeature.AddImsMsFeature(recessiveImsMsFeature);
}
}
recessiveFeature.imsMsFeatureList.Clear();
}
public static IEnumerable <LCIMSMSFeature> ClusterByMassAndScanLC(IEnumerable <imsMsFeature> imsMsFeatureEnumerable)
{
var lcimsmsFeatureList = new List <LCIMSMSFeature>();
var massToleranceBase = Settings.MassMonoisotopicConstraint;
var sortByMassQuery = from imsMsFeature in imsMsFeatureEnumerable
orderby imsMsFeature.CalculateAverageMonoisotopicMass()
select imsMsFeature;
LCIMSMSFeature lcimsmsFeature = null;
double massReference = -99;
foreach (var imsMsFeature in sortByMassQuery)
{
var mass = imsMsFeature.CalculateAverageMonoisotopicMass();
var massTolerance = massToleranceBase * massReference / 1000000;
var massToleranceHigh = massReference + massTolerance;
var massToleranceLow = massReference - massTolerance;
if (mass >= massToleranceLow && mass <= massToleranceHigh && lcimsmsFeature != null)
{
lcimsmsFeature.AddImsMsFeature(imsMsFeature);
}
else
{
lcimsmsFeature = new LCIMSMSFeature(imsMsFeature.Charge);
lcimsmsFeature.AddImsMsFeature(imsMsFeature);
lcimsmsFeatureList.Add(lcimsmsFeature);
}
massReference = mass;
}
return(lcimsmsFeatureList);
//IEnumerable<LCIMSMSFeature> splitLCIMSMSFeatureEnumerable = SplitByScanLCGap(lcimsmsFeatureList);
//return splitLCIMSMSFeatureEnumerable;
}
public static IEnumerable <LCIMSMSFeature> SplitLCIMSMSFeaturesByScanLC(IEnumerable <LCIMSMSFeature> lcImsMsFeatureEnumerable)
{
var lcimsmsFeatureList = new List <LCIMSMSFeature>();
var lcImsMsFeatureIndex = 0;
foreach (var lcimsmsFeature in lcImsMsFeatureEnumerable)
{
var sortByScanLC = from imsMsFeature in lcimsmsFeature.imsMsFeatureList
orderby imsMsFeature.ScanLC
select imsMsFeature;
LCIMSMSFeature newLcImsMsFeature = null;
var referenceScanLC = -99999;
foreach (var imsMsFeature in sortByScanLC)
{
var scanLC = imsMsFeature.ScanLC;
if (scanLC - referenceScanLC > Settings.LCGapSizeMax)
{
newLcImsMsFeature = new LCIMSMSFeature(imsMsFeature.Charge);
newLcImsMsFeature.AddImsMsFeature(imsMsFeature);
newLcImsMsFeature.OriginalIndex = lcImsMsFeatureIndex;
lcImsMsFeatureIndex++;
lcimsmsFeatureList.Add(newLcImsMsFeature);
}
else
{
newLcImsMsFeature?.AddImsMsFeature(imsMsFeature);
}
referenceScanLC = scanLC;
}
}
return(lcimsmsFeatureList);
}
public static IEnumerable <LCIMSMSFeature> FindDriftTimePeaks(Peak driftProfilePeak, LCIMSMSFeature lcimsmsFeature, double averageTOFLength, double framePressure)
{
var imsMsFeatureList = lcimsmsFeature.imsMsFeatureList;
var sortByScanLCQuery = from imsMsFeature in imsMsFeatureList
orderby imsMsFeature.ScanLC
select imsMsFeature;
var globalIMSScanMinimum = double.MaxValue;
var globalIMSScanMaximum = double.MinValue;
// Grab all of the intensity values for each IMS-MS Feature and find the global minimum and maximum Drift Times
foreach (var imsMsFeature in sortByScanLCQuery)
{
imsMsFeature.GetMinAndMaxIMSScan(out var localIMSScanMinimum, out var localIMSScanMaximum);
if (localIMSScanMinimum < globalIMSScanMinimum)
{
globalIMSScanMinimum = localIMSScanMinimum;
}
if (localIMSScanMaximum > globalIMSScanMaximum)
{
globalIMSScanMaximum = localIMSScanMaximum;
}
}
var smoothedDriftProfilePeak = PeakUtil.KDESmooth(driftProfilePeak, Settings.SmoothingStDev); // TODO: Find a good value. 0.15? Less smooth = more conformations!
var smoothedDriftProfileInterpolation = PeakUtil.GetLinearInterpolationMethod(smoothedDriftProfilePeak);
var xyPairList = new List <XYPair>();
var peakList = new List <Peak>();
var previousIntensity = double.MinValue;
var movingUp = true;
// lcimsmsFeature.GetMinAndMaxScanLC(out var minScanLC, out var maxScanLC);
var minimumIntensityToConsider = smoothedDriftProfilePeak.GetMaximumYValue() * 0.05;
//DisplayPeakXYData(smoothedDriftProfilePeak);
//Console.WriteLine("Global IMS Scan Min = " + globalIMSScanMinimum + "\tGlobal IMS Scan Max = " + globalIMSScanMaximum);
for (var i = globalIMSScanMinimum; i <= globalIMSScanMaximum; i += 1)
{
var imsScan = i;
var intensity = smoothedDriftProfileInterpolation.Interpolate(imsScan);
if (intensity > minimumIntensityToConsider)
{
//Console.WriteLine(imsScan + "\t" + intensity + "\t" + movingUp);
if (intensity > previousIntensity)
{
// End of Peak
if (!movingUp && xyPairList.Count > 0)
{
PadXYPairsWithZeros(ref xyPairList, 2);
//xyPairList = PadXYPairsWithZeros(xyPairList, imsScanMinimum, i - DRIFT_TIME_SLICE_WIDTH, 1);
var peak = new Peak(xyPairList);
if (peak.XYPairList.Count >= 7)
{
peakList.Add(peak);
}
// Start over with a new Peak
xyPairList.Clear();
movingUp = true;
}
}
else
{
movingUp = false;
}
var xyPair = new XYPair(imsScan, intensity);
xyPairList.Add(xyPair);
previousIntensity = intensity;
}
else
{
movingUp = false;
previousIntensity = 0;
}
}
// When you get to the end, end the last Peak, but only if it has a non-zero value
if (xyPairList.Any(xyPair => xyPair.YValue > minimumIntensityToConsider))
{
PadXYPairsWithZeros(ref xyPairList, 2);
//xyPairList = PadXYPairsWithZeros(xyPairList, imsScanMinimum, globalIMSScanMaximum, 1);
var lastPeak = new Peak(xyPairList);
if (lastPeak.XYPairList.Count >= 7)
{
peakList.Add(lastPeak);
}
}
// var resolvingPower = GetResolvingPower(lcimsmsFeature.Charge);
var newLcImsMsFeatureList = new List <LCIMSMSFeature>();
foreach (var peak in peakList)
{
var repIMSScan = peak.GetQuadraticFit();
// TODO: Fix this
//double theoreticalFWHM = driftTime / resolvingPower;
double theoreticalFWHM = 3;
peak.GetMinAndMaxXValues(out var minimumXValue, out var maximumXValue);
const int numPoints = 100;
var normalDistributionXYPairList = PeakUtil.CreateTheoreticalGaussianPeak(repIMSScan, theoreticalFWHM, numPoints);
PadXYPairsWithZeros(ref normalDistributionXYPairList, 5);
var normalDistributionPeak = new Peak(normalDistributionXYPairList);
var peakInterpolation = PeakUtil.GetLinearInterpolationMethod(peak);
var fitScore = PeakUtil.CalculatePeakFit(peak, normalDistributionPeak, 0);
// Create a new LC-IMS-MS Feature
var newLcImsMsFeature = new LCIMSMSFeature(lcimsmsFeature.Charge)
{
OriginalIndex = lcimsmsFeature.OriginalIndex,
IMSScore = (float)fitScore,
AbundanceMaxRaw = Math.Round(peak.GetMaximumYValue()),
// Using Math.Floor instead of Math.Round because I used to cast this to an int which is essentially Math.Floor.
// The difference is negligible, but OHSU would complain if results were the slightest bit different if the app was re-run on the same dataset.
AbundanceSumRaw = Math.Floor(peakInterpolation.Integrate(peak.GetMaximumXValue())),
DriftTime = ConvertIMSScanToDriftTime(repIMSScan, averageTOFLength, framePressure)
};
// Create new IMS-MS Features by grabbing MS Features in each LC Scan that are in the defined window of the detected drift time
foreach (var imsMsFeature in lcimsmsFeature.imsMsFeatureList)
{
var msFeatures = imsMsFeature.FindMSFeaturesInScanIMSRange(minimumXValue, maximumXValue).ToList();
if (!msFeatures.Any())
{
continue;
}
var newImsMsFeature = new imsMsFeature(imsMsFeature.ScanLC, imsMsFeature.Charge);
newImsMsFeature.AddMSFeatureList(msFeatures);
newLcImsMsFeature.AddImsMsFeature(newImsMsFeature);
}
if (newLcImsMsFeature.imsMsFeatureList.Count > 0)
{
newLcImsMsFeatureList.Add(newLcImsMsFeature);
/*
* // TODO: Find LC Peaks
* var sortByScanLC = from imsMsFeature in newLCIMSMSFeature.imsMsFeatureList
* orderby imsMsFeature.ScanLC ascending
* select imsMsFeature;
*
* Console.WriteLine("*************************************************");
* Console.WriteLine("Index = " + index + "\tMass = " + newLCIMSMSFeature.CalculateAverageMass() + "\tDrift = " + driftTime + "\tLC Range = " + sortByScanLC.First().ScanLC + "\t" + sortByScanLC.Last().ScanLC);
*
* List<XYPair> lcXYPairList = new List<XYPair>();
* int scanLC = sortByScanLC.First().ScanLC - 1;
*
* foreach (imsMsFeature imsMsFeature in sortByScanLC)
* {
* int currentScanLC = imsMsFeature.ScanLC;
*
* for (int i = scanLC + 1; i < currentScanLC; i++)
* {
* XYPair zeroValue = new XYPair(i, 0);
* lcXYPairList.Add(zeroValue);
* Console.Write("0\t");
* }
*
* XYPair xyPair = new XYPair(currentScanLC, imsMsFeature.GetIntensity());
* lcXYPairList.Add(xyPair);
*
* scanLC = currentScanLC;
*
* Console.Write(imsMsFeature.GetIntensity() + "\t");
* }
* Console.WriteLine(string.Empty);
* Console.WriteLine("*************************************************");
*/
// TODO: Calculate LC Score
}
else
{
//Console.WriteLine("$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$ FOUND EMPTY $$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$");
// TODO: Figure out why this actually happens. I believe that this SHOULD NOT happen. Below is a hack to return a conformation even if this happens
// It actually looks like most of these occurrences are due to large gaps in the drift time, which cause a small peak to be found in the gap which has no members.
//Console.WriteLine("**********************************************************************");
//Console.WriteLine("Detected Drift Time = " + driftTime + "\tLow = " + lowDriftTime + "\tHigh = " + highDriftTime);
//lcimsmsFeature.PrintLCAndDriftTimeMap();
//Console.WriteLine("**********************************************************************");
//Console.WriteLine("===============================================================");
//Console.WriteLine("DT = " + driftTime + "\tLow DT = " + lowDriftTime + "\tHigh DT = " + highDriftTime);
//Console.WriteLine("Global Min = " + globalDriftTimeMinimum + "\tGlobal Max = " + globalDriftTimeMaximum);
//peak.PrintPeakToConsole();
//Console.WriteLine("===============================================================");
}
}
// Find the Conformation that has the highest member count and store the value into all conformations of this LC-IMS-MS Feature
if (newLcImsMsFeatureList.Count > 0)
{
var maxMemberCount = newLcImsMsFeatureList.Select(feature => feature.GetMemberCount()).Max();
foreach (var feature in newLcImsMsFeatureList)
{
feature.MaxMemberCount = maxMemberCount;
}
}
return(newLcImsMsFeatureList);
}