/// <summary>
/// Generates alignment functions for alignment between features in each separation dimension.
/// Warps the elution value for each feature based on the alignment function.
/// </summary>
/// <param name="aligneeFeatures">The features to warp.</param>
/// <param name="baselineFeatures">The features to warp to.</param>
/// <param name="includeMassInMatchScore">
/// Should mass be considered when scoring a match between an alignee feature and baseline feature?
/// </param>
/// <returns>The list of alignment results for each separation dimension.</returns>
public NewAlignmentData WarpNet(List <UMCLight> aligneeFeatures, List <UMCLight> baselineFeatures, bool includeMassInMatchScore)
{
// Generate candidate matches: Match alignee features -> baseline features by mass only
var featureMatcher = new LcmsWarpFeatureMatcher(this.options);
featureMatcher.GenerateCandidateMatches(aligneeFeatures, baselineFeatures);
var warpedFeatures = new List <UMCLight>(aligneeFeatures);
var results = new Dictionary <FeatureLight.SeparationTypes, LcmsWarpResults>();
// Perform warp on each separation dimension.
foreach (var separationType in this.options.SeparationTypes)
{
// Get matches for given separation dimension
var matches = featureMatcher.GetMatchesAs(separationType);
// Calculate two dimensional statistics for mass and the current separation dimension.
var statistics = LcmsWarpStatistics.CalculateAndGetStatistics(matches);
// Calculate alignee sections
var aligneeSections = new LcmsWarpSectionInfo(this.options.NumTimeSections);
aligneeSections.InitSections(aligneeFeatures);
// Calculate baseline sections
var baselineSections = new LcmsWarpSectionInfo(this.options.NumTimeSections * this.options.ContractionFactor);
baselineSections.InitSections(baselineFeatures);
// Generate alignment function, only score sections based on NET.
var alignmentScorer = new LcmsWarpAlignmentScorer(this.options, includeMassInMatchScore, statistics);
var alignmentFunction = alignmentScorer.GetAlignment(aligneeSections, baselineSections, matches);
alignmentFunction.SeparationType = separationType;
// Create alignment score heatmap
var alignmentScoreHeatMap = LcmsWarpPlotDataCreator.GetAlignmentHeatMap(
alignmentScorer.AlignmentScoreMatrix,
true,
this.options.NumTimeSections,
this.options.NumBaselineSections,
this.options.MaxExpansionWidth);
// Warp the values in the features for this separation type
warpedFeatures = alignmentFunction.GetWarpedFeatures(warpedFeatures).ToList();
var dimensionResults = new LcmsWarpResults
{
AlignmentFunction = alignmentFunction,
Statistics = statistics,
AlignmentScoreHeatMap = alignmentScoreHeatMap
};
results.Add(separationType, dimensionResults);
}
return(new NewAlignmentData
{
SeparationAlignments = results,
AlignedFeatures = warpedFeatures,
});
}
/// <summary>
/// Calculates the Standard deviations of the matches.
/// Note: method requires more than 6 matches to produce meaningful results.
/// </summary>
public static LcmsWarpStatistics CalculateAndGetStatistics(List <LcmsWarpFeatureMatch> matches)
{
if (matches.Count <= RequiredMatches)
{
throw new ArgumentException(string.Format("This requires at least {0} matches to produce meaningful results", RequiredMatches));
}
// Calculate NET delta (difference between alignee and baseline NET).
var massDeltas = new List <double>(matches.Count);
var netDeltas = new List <double>(matches.Count);
foreach (var match in matches)
{
var baselineFeature = match.BaselineFeature;
massDeltas.Add(((baselineFeature.MassMonoisotopic - match.AligneeFeature.MassMonoisotopic) * 1000000) /
match.AligneeFeature.MassMonoisotopic);
netDeltas.Add(match.BaselineNet - match.Net);
}
// Two dimensional expectation maximization
double normalProbability, falseHitProbDensity, muMass, muNet, massStdDev, netStdDev;
MathUtilities.TwoDem(
massDeltas,
netDeltas,
out normalProbability,
out falseHitProbDensity,
out muMass,
out muNet,
out massStdDev,
out netStdDev);
var statistics = new LcmsWarpStatistics
{
MassStdDev = massStdDev,
NetStdDev = netStdDev,
FalseHitProbDensity = falseHitProbDensity,
NormalProbability = normalProbability
};
return(statistics);
}