public void TestLoess(FitFunctionTypes type,
FitFunctionTypes generatingFunction,
double dt,
double attenuation)
{
var interpolator = new LoessInterpolator(.25, 0);
var xValues = new List <double>();
var yValues = new List <double>();
double cv = 0;
var fitFunction = FitFunctionFactory.Create(type);
var genFunction = FitFunctionFactory.Create(generatingFunction);
var random = new Random();
// Take one period of the sine wave...
while (cv < Math.PI * 3)
{
var value = genFunction(cv) + random.NextDouble() * attenuation;
xValues.Add(cv);
yValues.Add(value);
cv += dt;
}
var newYValues = interpolator.Smooth(xValues, yValues, fitFunction);
for (var i = 0; i < xValues.Count; i++)
{
Console.WriteLine(@"{0} {1} {2} {3}", i, xValues[i], yValues[i], newYValues[i]);
}
dt /= 2;
cv = 0;
Console.WriteLine();
// Take one period of the sine wave...
while (cv < Math.PI * 3)
{
var predicted = interpolator.Predict(cv);
Console.WriteLine(@"{0} {1}", cv, predicted);
cv += dt;
}
}
private static void WriteMatches(
string name,
ISpectralAnalysisWriter writer,
List <double> xvalues,
List <double> yvalues,
IList <double> fit,
LoessInterpolator interpolator,
List <SpectralAnchorPointMatch> anchorPoints,
List <double> preNet,
List <double> postNet,
Func <double, double, double> difference)
{
if (anchorPoints == null)
{
throw new ArgumentNullException("anchorPoints");
}
writer.WriteLine(string.Format(@"[{0}]", name));
writer.WriteLine("x\ty\tfit\tpre-diff\tpost-diff\tsim score\tvalid");
for (var index = 0; index < xvalues.Count; index++)
{
var x = xvalues[index];
var y = yvalues[index];
var value = interpolator.Predict(x);
var preDiff = difference(x, y);
var postDiff = difference(value, y);
writer.WriteLine(string.Format("{0}\t{1}\t{2}\t{3}\t{4}\t{5}\t{6}",
x,
y,
fit[index],
preDiff,
postDiff,
anchorPoints[index].SimilarityScore,
anchorPoints[index].IsValidMatch));
preNet.Add(preDiff);
postNet.Add(postDiff);
}
}
public void TestAlignmentfunction(
string path,
string outputPath,
double bandwidth,
int robustnessIterators
)
{
var data = File.ReadAllLines(path);
var x = new List <double>();
var y = new List <double>();
for (var i = 1; i < data.Length; i++)
{
var columns = data[i].Split('\t');
if (columns.Count() < 4)
{
continue;
}
x.Add(Convert.ToDouble(columns[0]));
y.Add(Convert.ToDouble(columns[2]));
}
using (var writer = File.CreateText(outputPath))
{
var loess = new LoessInterpolator(bandwidth, robustnessIterators);
loess.MaxDistance = bandwidth;
loess.Smooth(x, y, FitFunctionFactory.Create(FitFunctionTypes.Cubic));
writer.WriteLine("NET\tNETY\tAligned\tNET\tERRORY\tERROR-Aligned");
for (var i = 0; i < y.Count; i++)
{
var value = loess.Predict(y[i]);
writer.WriteLine("{0}\t{1}\t{2}\t{3}\t{4}\t{5}", x[i], y[i], value, x[i], y[i] - x[i], value - x[i]);
}
}
}
private static void WriteMatches(
string name,
ISpectralAnalysisWriter writer,
List<double> xvalues,
List<double> yvalues,
IList<double> fit,
LoessInterpolator interpolator,
List<SpectralAnchorPointMatch> anchorPoints,
List<double> preNet,
List<double> postNet,
Func<double, double, double> difference)
{
if (anchorPoints == null) throw new ArgumentNullException("anchorPoints");
writer.WriteLine(string.Format(@"[{0}]", name));
writer.WriteLine("x\ty\tfit\tpre-diff\tpost-diff\tsim score\tvalid");
for (var index = 0; index < xvalues.Count; index++)
{
var x = xvalues[index];
var y = yvalues[index];
var value = interpolator.Predict(x);
var preDiff = difference(x, y);
var postDiff = difference(value, y);
writer.WriteLine(string.Format("{0}\t{1}\t{2}\t{3}\t{4}\t{5}\t{6}",
x,
y,
fit[index],
preDiff,
postDiff,
anchorPoints[index].SimilarityScore,
anchorPoints[index].IsValidMatch));
preNet.Add(preDiff);
postNet.Add(postDiff);
}
}
public void TestLoess(FitFunctionTypes type,
FitFunctionTypes generatingFunction,
double dt,
double attenuation)
{
var interpolator = new LoessInterpolator(.25, 0);
var xValues = new List<double>();
var yValues = new List<double>();
double cv = 0;
var fitFunction = FitFunctionFactory.Create(type);
var genFunction = FitFunctionFactory.Create(generatingFunction);
var random = new Random();
// Take one period of the sine wave...
while (cv < Math.PI*3)
{
var value = genFunction(cv) + random.NextDouble()*attenuation;
xValues.Add(cv);
yValues.Add(value);
cv += dt;
}
var newYValues = interpolator.Smooth(xValues, yValues, fitFunction);
for (var i = 0; i < xValues.Count; i++)
{
Console.WriteLine(@"{0} {1} {2} {3}", i, xValues[i], yValues[i], newYValues[i]);
}
dt /= 2;
cv = 0;
Console.WriteLine();
// Take one period of the sine wave...
while (cv < Math.PI*3)
{
var predicted = interpolator.Predict(cv);
Console.WriteLine(@"{0} {1}", cv, predicted);
cv += dt;
}
}
public void CreateAlignmentFunctions(IEnumerable <SpectralAnchorPointMatch> matches)
{
var netXvalues = new List <double>();
var netYvalues = new List <double>();
var massXvalues = new List <double>();
var massYvalues = new List <double>();
matches = matches.ToList().OrderBy(x => x.AnchorPointX.Net);
// 1. Find the best matches
// 2. Find only matches that have been made once.
var bestMatches = new Dictionary <int, SpectralAnchorPointMatch>();
foreach (var match in matches)
{
var scan = match.AnchorPointX.Scan;
if (bestMatches.ContainsKey(scan))
{
if (bestMatches[scan].SimilarityScore < match.SimilarityScore)
{
bestMatches[scan] = match;
}
}
else
{
bestMatches.Add(scan, match);
}
}
// 2. Find only those matched once
var all = new Dictionary <int, SpectralAnchorPointMatch>();
foreach (var match in bestMatches.Values)
{
var scan = match.AnchorPointY.Scan;
if (all.ContainsKey(scan))
{
if (all[scan].SimilarityScore < match.SimilarityScore)
{
all[scan] = match;
}
}
else
{
all.Add(scan, match);
}
}
// Then generate the NET Alignment using R1
var anchorPoints = all.Values.OrderBy(x => x.AnchorPointY.Net).ToList();
matches =
anchorPoints.Where(
x => FeatureLight.ComputeMassPPMDifference(x.AnchorPointX.Mz, x.AnchorPointY.Mz) < 20 &&
x.AnchorPointX.Spectrum.ParentFeature.ChargeState == x.AnchorPointY.Spectrum.ParentFeature.ChargeState
).ToList();
foreach (var match in matches)
{
netXvalues.Add(match.AnchorPointX.Net);
netYvalues.Add(match.AnchorPointY.Net);
}
var netInterpolator = new LoessInterpolator(Bandwidth, 5);
netInterpolator.Smooth(netYvalues, netXvalues, FitFunctionFactory.Create(FitFunctionTypes.TriCubic));
// Then generate the Mass Alignment using R1
// We also have to resort the matches based on mass now too
anchorPoints = all.Values.OrderBy(x => x.AnchorPointY.Mz).ToList();
foreach (var match in anchorPoints)
{
massXvalues.Add(match.AnchorPointX.Mz);
massYvalues.Add(match.AnchorPointY.Mz);
}
var massInterpolator = new LoessInterpolator();
massInterpolator.Smooth(massYvalues, massXvalues, FitFunctionFactory.Create(FitFunctionTypes.TriCubic));
m_netInterpolator = netInterpolator;
m_massInterpolator = massInterpolator;
foreach (var match in anchorPoints)
{
match.AnchorPointY.NetAligned = netInterpolator.Predict(match.AnchorPointY.Net);
match.AnchorPointY.MzAligned = massInterpolator.Predict(match.AnchorPointY.Mz);
}
}
public double AlignNet(double net)
{
return(m_netInterpolator.Predict(net));
}
public double AlignMass(double mass)
{
return(m_massInterpolator.Predict(mass));
}
public void TestAlignmentfunction(
string path,
string outputPath,
double bandwidth,
int robustnessIterators
)
{
var data = File.ReadAllLines(path);
var x = new List<double>();
var y = new List<double>();
for (var i = 1; i < data.Length; i++)
{
var columns = data[i].Split('\t');
if (columns.Count() < 4)
continue;
x.Add(Convert.ToDouble(columns[0]));
y.Add(Convert.ToDouble(columns[2]));
}
using (var writer = File.CreateText(outputPath))
{
var loess = new LoessInterpolator(bandwidth, robustnessIterators);
loess.MaxDistance = bandwidth;
loess.Smooth(x, y, FitFunctionFactory.Create(FitFunctionTypes.Cubic));
writer.WriteLine("NET\tNETY\tAligned\tNET\tERRORY\tERROR-Aligned");
for (var i = 0; i < y.Count; i++)
{
var value = loess.Predict(y[i]);
writer.WriteLine("{0}\t{1}\t{2}\t{3}\t{4}\t{5}", x[i], y[i], value, x[i], y[i] - x[i], value - x[i]);
}
}
}