private void PlotFit(PoolParameters poolParams, string label, double kbi = 1000000) //default kbi is just a very high number to simulate instantaneous turnover (proxy for the available amino acid pool)
{
double[] timepoints = new double[1000];
for (int i = 0; i < timepoints.Length; i++)
{
timepoints[i] = i / 10.0;
}
//half-life = ln(2)/kbt, make half life 0, kbt = infinity
double[] rfs = NonLinearRegression.PredictRelativeFractionUsingThreeCompartmentModel(poolParams.Kst, poolParams.Kbt, poolParams.Kao, kbi, timepoints);
Dispatcher.Invoke(() =>
{
RatioComparisonPlot.plt.Layout(titleHeight: 20, xLabelHeight: 40, y2LabelWidth: 20);
RatioComparisonPlot.plt.XLabel("Time (Days)", fontSize: 20);
// RatioComparisonPlot.plt.YLabel("Relative Fraction (Lys0/Total)", fontSize: 20);
RatioComparisonPlot.plt.YLabel("Lys0 / LysTotal", fontSize: 20);
RatioComparisonPlot.plt.Axis(0, 100, 0, 1);
RatioComparisonPlot.plt.Ticks(fontSize: 18);
RatioComparisonPlot.plt.PlotScatter(timepoints, rfs, label: label, markerShape: ScottPlot.MarkerShape.none);
if (DisplayLegendCheckBox.IsChecked.Value)
{
RatioComparisonPlot.plt.Legend();
}
RatioComparisonPlot.Render();
});
}
private void PlotPrecisionScatterPlot(List <PeptideTurnoverObject> peptidesToPlot, PoolParameters poolParams)
{
PrecisionPlot.plt.Clear();
PrecisionPlot.plt.GetPlottables().Clear();
if (peptidesToPlot.Count == 0)
{
return;
}
Dictionary <double, List <(double halfLife, double relativeFraction)> > dictionaryToPlot = new Dictionary <double, List <(double halfLife, double relativeFraction)> >();
foreach (PeptideTurnoverObject peptide in peptidesToPlot)
{
//grab measurements
double halfLife = Math.Log(2, Math.E) / peptide.Kbi;
for (int i = 0; i < peptide.Timepoints.Length; i++)
{
if (dictionaryToPlot.ContainsKey(peptide.Timepoints[i]))
{
dictionaryToPlot[peptide.Timepoints[i]].Add((halfLife, peptide.RelativeFractions[i]));
}
else
{
dictionaryToPlot[peptide.Timepoints[i]] = new List <(double halfLife, double relativeFraction)> {
(halfLife, peptide.RelativeFractions[i])
};
}
}
}
//plot all peptide data
double[] timepoints = dictionaryToPlot.Keys.OrderBy(x => x).ToArray();
foreach (double timepoint in timepoints)
{
var value = dictionaryToPlot[timepoint];
PrecisionPlot.plt.PlotScatter(value.Select(x => x.halfLife).ToArray(), value.Select(x => x.relativeFraction).ToArray(), lineWidth: 0, markerSize: 3, label: timepoint.ToString(), markerShape: ScottPlot.MarkerShape.openCircle);
}
//plt fits for each timepoint on top of the peptide data
double[] halflives = new double[2499];
for (int i = 0; i < halflives.Length; i++)
{
halflives[i] = i / 5.0 + 0.2;
}
List <double>[] rfs = new List <double> [timepoints.Length];
for (int i = 0; i < timepoints.Length; i++)
{
rfs[i] = new List <double>();
}
foreach (double halflife in halflives)
{
if (halflife == 0)
{
continue;
}
//halflife = ln(2)/kbi
//kbi = ln(2)/halflife
double[] rfsForThisHalfLife = NonLinearRegression.PredictRelativeFractionUsingThreeCompartmentModel(
poolParams.Kst, poolParams.Kbt, poolParams.Kao, Math.Log(2, Math.E) / halflife, timepoints);
for (int i = 0; i < timepoints.Length; i++)
{
rfs[i].Add(rfsForThisHalfLife[i]);
}
}
PrecisionPlot.plt.Axis(0, 50, 0, 1);
for (int i = 0; i < timepoints.Length; i++)
{
PrecisionPlot.plt.PlotScatter(halflives, rfs[i].ToArray(), Color.Black, markerSize: 0);
}
if (DisplayLegendCheckBox.IsChecked.Value)
{
PrecisionPlot.plt.Legend(location: ScottPlot.legendLocation.upperRight);
}
else
{
PrecisionPlot.plt.Legend(false);
}
PrecisionPlot.plt.YLabel("Lys0 / LysTotal");
PrecisionPlot.plt.XLabel("Half-life (Days)");
PrecisionPlot.plt.Axis(0, 50, 0, 1);
PrecisionPlot.Render(); PrecisionPlot.Render();
}
public static void CompareProteinsAcrossFiles(List <string> filenames, List <PeptideTurnoverObject> allProteins, Dictionary <string, PoolParameters> poolParameterDictionary)
{
if (filenames.Count < 2)
{
return;
}
string outputDirectory = Path.GetDirectoryName(filenames.First());
Directory.CreateDirectory(Path.Combine(outputDirectory, "StatisticalComparisons"));
for (int i = 0; i < filenames.Count; i++)
{
string fileOne = filenames[i];
PoolParameters paramsOne = poolParameterDictionary[fileOne];
//get the proteins for this file
List <PeptideTurnoverObject> proteinsForFileOne = allProteins.Where(x => fileOne.Equals(x.FileName)).OrderBy(x => x.Protein).ToList();
for (int j = i + 1; j < filenames.Count; j++)
{
string fileTwo = filenames[j];
PoolParameters paramsTwo = poolParameterDictionary[fileTwo];
List <string> linesToWrite = new List <string>();
//add header
linesToWrite.Add("Protein\tFold Change\tNeg. log(p-Value)\tHalf-life " + fileOne + "\tHalf-life " + fileTwo);
List <PeptideTurnoverObject> proteinsForFileTwo = allProteins.Where(x => fileTwo.Equals(x.FileName)).OrderBy(x => x.Protein).ToList();
//get the overlap between them
int a = 0;
int b = 0;
while (a < proteinsForFileOne.Count && b < proteinsForFileTwo.Count)
{
PeptideTurnoverObject proteinOne = proteinsForFileOne[a];
PeptideTurnoverObject proteinTwo = proteinsForFileTwo[b];
int comparison = (proteinOne.Protein).CompareTo(proteinTwo.Protein);
if (comparison == 0)
{
//do the comparison (t-test of montecarlos, which dramatically overestimates the sample size)
//Sample sampleOne = new Sample(proteinOne.MonteCarloKbis.Select(x => Math.Log10(2) / x));
//Sample sampleTwo = new Sample(proteinTwo.MonteCarloKbis.Select(x => Math.Log10(2) / x));
//TestResult result = Sample.StudentTTest(sampleOne, sampleTwo);
//linesToWrite.Add(proteinOne.Protein + "\t" + (Math.Log2(sampleTwo.Median) - Math.Log2(sampleOne.Median)).ToString() + '\t' +
// (-1*Math.Log10(result.Probability)).ToString() + '\t' + (Math.Log10(2) / proteinOne.Kbi).ToString() + '\t' + (Math.Log10(2) / proteinTwo.Kbi).ToString());
//do the comparison (t-test of normalized ratios for all timepoints)
double averageKbi = (proteinOne.Kbi + proteinTwo.Kbi) / 2;
double normalizedHalfLife = Math.Log(2) / (averageKbi); //this is the day we're going to normalize all of the relative fractions to
//create an array of a single value (the normalized timepoint) to create a new timepoint array
double[] comparisonTimepointsOne = new double[proteinOne.Timepoints.Length];
double[] comparisonTimepointsTwo = new double[proteinTwo.Timepoints.Length];
for (int index = 0; index < comparisonTimepointsOne.Length; index++)
{
comparisonTimepointsOne[index] = normalizedHalfLife;
}
for (int index = 0; index < comparisonTimepointsTwo.Length; index++)
{
comparisonTimepointsTwo[index] = normalizedHalfLife;
}
//predict the expected values for the ratios of protein one based on the fit of the comparison
double[] expectedOriginalRatiosOne = NonLinearRegression.PredictRelativeFractionUsingThreeCompartmentModel(paramsOne.Kst, paramsOne.Kbt, paramsOne.Kao, averageKbi, proteinOne.Timepoints);
//predict the expected values for the ratios of proteoform one based on the fit of the comparison if they were all at the same normalized timepoint
double[] expectedUpdatedRatiosOne = NonLinearRegression.PredictRelativeFractionUsingThreeCompartmentModel(paramsOne.Kst, paramsOne.Kbt, paramsOne.Kao, averageKbi, comparisonTimepointsOne);
//do the same thing with protein two
double[] expectedOriginalRatiosTwo = NonLinearRegression.PredictRelativeFractionUsingThreeCompartmentModel(paramsTwo.Kst, paramsTwo.Kbt, paramsTwo.Kao, averageKbi, proteinTwo.Timepoints);
double[] expectedUpdatedRatiosTwo = NonLinearRegression.PredictRelativeFractionUsingThreeCompartmentModel(paramsTwo.Kst, paramsTwo.Kbt, paramsTwo.Kao, averageKbi, comparisonTimepointsTwo);
//create empty arrays for the normalized ratios
double[] normalizedRatiosOne = new double[expectedOriginalRatiosOne.Length];
double[] normalizedRatiosTwo = new double[expectedOriginalRatiosTwo.Length];
//calculate the normalized ratios by subtracting the expected ratio (so that we are measuring the residual between the point and the comparison fit) and then adding the ratio of the comparison fit at the normalized timepoint.
for (int index = 0; index < proteinOne.RelativeFractions.Length; index++)
{
//the normalized ratio is equal to the original ratio minus the original fit to the data plus the fit if the kbi was averaged
normalizedRatiosOne[index] = proteinOne.RelativeFractions[index] - expectedOriginalRatiosOne[index] + expectedUpdatedRatiosOne[index];
}
for (int index = 0; index < proteinTwo.RelativeFractions.Length; index++)
{
normalizedRatiosTwo[index] = proteinTwo.RelativeFractions[index] - expectedOriginalRatiosTwo[index] + expectedUpdatedRatiosTwo[index];
}
Sample sampleOne = new Sample(normalizedRatiosOne);
Sample sampleTwo = new Sample(normalizedRatiosTwo);
TestResult result = Sample.StudentTTest(sampleOne, sampleTwo);
linesToWrite.Add(proteinOne.Protein + "\t" + (Math.Log2(Math.Log(2) / proteinTwo.Kbi) - Math.Log2(Math.Log(2) / proteinOne.Kbi)).ToString() + '\t' +
(-1 * Math.Log10(result.Probability)).ToString() + '\t' + (Math.Log(2) / proteinOne.Kbi).ToString() + '\t' + (Math.Log(2) / proteinTwo.Kbi).ToString());
a++;
b++;
}
else if (comparison < 0)
{
a++;
}
else
{
b++;
}
}
File.WriteAllLines(Path.Combine(outputDirectory, "StatisticalComparisons", "Comparison_" + Path.GetFileNameWithoutExtension(fileOne) + "vs" + Path.GetFileNameWithoutExtension(fileTwo) + ".tsv"), linesToWrite);
}
}
}
public static void CompareProteoformsWithinFiles(List <string> filenames, List <PeptideTurnoverObject> allProteins, Dictionary <string, PoolParameters> poolParameterDictionary)
{
for (int fileIndex = 0; fileIndex < filenames.Count; fileIndex++)
{
string filename = filenames[fileIndex];
PoolParameters poolParams = poolParameterDictionary[filename];
List <PeptideTurnoverObject> proteinsForThisFile = allProteins.Where(x => filename.Equals(x.FileName)).OrderBy(x => x.Proteoform).ToList();
List <string> linesToWrite = new List <string>();
linesToWrite.Add("Proteoform A\tProteoform B\tHalf-life A\tHalf-life B\tLog2(Fold Change)\tNeg. log(p-Value)");
int indexOfNextProteoformFamily = 0;
for (int i = 0; i < proteinsForThisFile.Count; i++)
{
string currentProtein = proteinsForThisFile[i].Proteoform.Split('_')[0];
//find last index for this proteoform family
indexOfNextProteoformFamily++;
for (; indexOfNextProteoformFamily < proteinsForThisFile.Count; indexOfNextProteoformFamily++)
{
if (!currentProtein.Equals(proteinsForThisFile[indexOfNextProteoformFamily].Proteoform.Split('_')[0]))
{
break;
}
}
for (; i < indexOfNextProteoformFamily; i++)
{
PeptideTurnoverObject proteinOne = proteinsForThisFile[i];
//see if it has a localized mod (or localized unmodified site), otherwise skip
string[] proteoformOne = proteinOne.Proteoform.Split('@').ToArray();
if (proteoformOne.Length == 2)
{
for (int j = i + 1; j < indexOfNextProteoformFamily; j++)
{
PeptideTurnoverObject proteinTwo = proteinsForThisFile[j];
string[] proteoformTwo = proteinTwo.Proteoform.Split('@').ToArray();
//if these are a pair for the same modification site, then do the comparison
if (proteoformTwo.Length == 2 && proteoformOne[1].Equals(proteoformTwo[1]))
{
//do the comparison (t-test of normalized ratios for all timepoints)
double averageKbi = (proteinOne.Kbi + proteinTwo.Kbi) / 2;
double normalizedHalfLife = Math.Log(2) / (averageKbi); //this is the day we're going to normalize all of the relative fractions to
//create an array of a single value (the normalized timepoint) to create a new timepoint array
double[] comparisonTimepointsOne = new double[proteinOne.Timepoints.Length];
double[] comparisonTimepointsTwo = new double[proteinTwo.Timepoints.Length];
for (int index = 0; index < comparisonTimepointsOne.Length; index++)
{
comparisonTimepointsOne[index] = normalizedHalfLife;
}
for (int index = 0; index < comparisonTimepointsTwo.Length; index++)
{
comparisonTimepointsTwo[index] = normalizedHalfLife;
}
double[] expectedOriginalRatiosOne = NonLinearRegression.PredictRelativeFractionUsingThreeCompartmentModel(poolParams.Kst, poolParams.Kbt, poolParams.Kao, averageKbi, proteinOne.Timepoints);
double[] expectedUpdatedRatiosOne = NonLinearRegression.PredictRelativeFractionUsingThreeCompartmentModel(poolParams.Kst, poolParams.Kbt, poolParams.Kao, averageKbi, comparisonTimepointsOne);
double[] expectedOriginalRatiosTwo = NonLinearRegression.PredictRelativeFractionUsingThreeCompartmentModel(poolParams.Kst, poolParams.Kbt, poolParams.Kao, averageKbi, proteinTwo.Timepoints);
double[] expectedUpdatedRatiosTwo = NonLinearRegression.PredictRelativeFractionUsingThreeCompartmentModel(poolParams.Kst, poolParams.Kbt, poolParams.Kao, averageKbi, comparisonTimepointsTwo);
double[] normalizedRatiosOne = new double[expectedOriginalRatiosOne.Length];
double[] normalizedRatiosTwo = new double[expectedOriginalRatiosTwo.Length];
for (int index = 0; index < proteinOne.RelativeFractions.Length; index++)
{
//the normalized ratio is equal to the original ratio minus the original fit to the data plus the fit if the kbi was averaged
normalizedRatiosOne[index] = proteinOne.RelativeFractions[index] - expectedOriginalRatiosOne[index] + expectedUpdatedRatiosOne[index];
}
for (int index = 0; index < proteinTwo.RelativeFractions.Length; index++)
{
normalizedRatiosTwo[index] = proteinTwo.RelativeFractions[index] - expectedOriginalRatiosTwo[index] + expectedUpdatedRatiosTwo[index];
}
Sample sampleOne = new Sample(normalizedRatiosOne);
Sample sampleTwo = new Sample(normalizedRatiosTwo);
TestResult result = Sample.StudentTTest(sampleOne, sampleTwo);
try //sometimes crashes if stdev is zero
{
linesToWrite.Add(proteinOne.Proteoform + "\t" + proteinTwo.Proteoform + '\t' + (Math.Log(2) / proteinOne.Kbi).ToString() + '\t' + (Math.Log(2) / proteinTwo.Kbi).ToString() + '\t' +
(Math.Log2((Math.Log(2) / proteinTwo.Kbi)) - Math.Log2((Math.Log(2) / proteinOne.Kbi))).ToString() + '\t' + (-1 * Math.Log(result.Probability)).ToString());
}
catch
{
linesToWrite.Add(proteinOne.Proteoform + "\t" + proteinTwo.Proteoform + '\t' + (Math.Log(2) / proteinOne.Kbi).ToString() + '\t' + (Math.Log(2) / proteinTwo.Kbi).ToString() + '\t' +
(Math.Log2(sampleTwo.Median) - Math.Log2(sampleOne.Median)).ToString() + '\t' + "NA");
}
}
}
}
}
i--;
}
File.WriteAllLines(Path.Combine(Path.GetDirectoryName(filename), Path.GetFileNameWithoutExtension(filename) + "_Results", Path.GetFileNameWithoutExtension(filename) + "_ProteoformAnalysis.tsv"), linesToWrite);
}
}