public static bool TryGetRegressionLine(IList<double> listIndependent, IList<double> listDependent, int minPoints, out RegressionLine line)
{
line = null;
if (listIndependent.Count != listDependent.Count || listIndependent.Count < minPoints)
return false;
double correlation;
while (true)
{
var statIndependent = new Statistics(listIndependent);
var statDependent = new Statistics(listDependent);
line = new RegressionLine(statDependent.Slope(statIndependent), statDependent.Intercept(statIndependent));
correlation = statDependent.R(statIndependent);
if (correlation >= MIN_IRT_TO_TIME_CORRELATION || listIndependent.Count <= minPoints)
break;
var furthest = 0;
var maxDistance = 0.0;
for (var i = 0; i < listDependent.Count; i++)
{
var distance = Math.Abs(line.GetY(listDependent[i]) - listIndependent[i]);
if (distance > maxDistance)
{
furthest = i;
maxDistance = distance;
}
}
listIndependent.RemoveAt(furthest);
listDependent.RemoveAt(furthest);
}
return correlation >= MIN_IRT_TO_TIME_CORRELATION;
}
public RegressionGraphPane(RegressionGraphData graphData)
{
_graphData = graphData;
Title.Text = graphData.Title;
XAxis.Title.Text = graphData.LabelX;
YAxis.Title.Text = graphData.LabelY;
Border.IsVisible = false;
Title.IsVisible = true;
Chart.Border.IsVisible = false;
XAxis.Scale.MaxAuto = true;
XAxis.Scale.MinAuto = true;
YAxis.Scale.MaxAuto = true;
YAxis.Scale.MinAuto = true;
Y2Axis.IsVisible = false;
X2Axis.IsVisible = false;
XAxis.MajorTic.IsOpposite = false;
YAxis.MajorTic.IsOpposite = false;
XAxis.MinorTic.IsOpposite = false;
YAxis.MinorTic.IsOpposite = false;
IsFontsScaled = false;
YAxis.Scale.MaxGrace = 0.1;
// Legend.FontSpec.Size = 12;
var curve = AddCurve(Resources.RegressionGraphPane_RegressionGraphPane_Values, graphData.XValues, graphData.YValues,
Color.Black, SymbolType.Diamond);
curve.Line.IsVisible = false;
curve.Symbol.Border.IsVisible = false;
curve.Symbol.Fill = new Fill(COLOR_REGRESSION);
// Find maximum points for drawing the regression line
var lineX = new[] { double.MaxValue, double.MinValue };
var lineY = new[] { double.MaxValue, double.MinValue };
for (int i = 0; i < graphData.XValues.Length; i++)
{
double xValue = graphData.XValues[i];
double yValue = graphData.YValues[i];
if (xValue < lineX[0])
{
lineX[0] = xValue;
lineY[0] = yValue;
}
if (xValue > lineX[1])
{
lineX[1] = xValue;
lineY[1] = yValue;
}
}
if (graphData.RegressionLine != null)
{
// Recalculate the y values based on the maximum x values
// and the regression.
lineY[0] = graphData.RegressionLine.GetY(lineX[0]);
lineY[1] = graphData.RegressionLine.GetY(lineX[1]);
curve = AddCurve(Resources.RegressionGraphPane_RegressionGraphPane_Regression, lineX, lineY, COLOR_LINE_REGRESSION);
curve.Line.IsAntiAlias = true;
curve.Line.IsOptimizedDraw = true;
Statistics statsX = new Statistics(_graphData.XValues);
Statistics statsY = new Statistics(_graphData.YValues);
double slope = statsY.Slope(statsX);
double intercept = statsY.Intercept(statsX);
_labelRegression = string.Format("{0} = {1:F04}, {2} = {3:F04}\n" + "r = {4:F02}", // Not L10N
Resources.Regression_slope,
slope,
Resources.Regression_intercept,
intercept,
statsY.R(statsX));
}
var regressionLineCurrent = graphData.RegressionLineCurrent;
if (regressionLineCurrent != null)
{
lineY[0] = regressionLineCurrent.GetY(lineX[0]);
lineY[1] = regressionLineCurrent.GetY(lineX[1]);
curve = AddCurve(Resources.RegressionGraphPane_RegressionGraphPane_Current, lineX, lineY, COLOR_LINE_REGRESSION_CURRENT);
curve.Line.IsAntiAlias = true;
curve.Line.IsOptimizedDraw = true;
curve.Line.Style = DashStyle.Dash;
_labelRegressionCurrent = string.Format("{0} = {1:F04}, {2} = {3:F04}", // Not L10N
Resources.Regression_slope,
regressionLineCurrent.Slope,
Resources.Regression_intercept,
regressionLineCurrent.Intercept);
}
}
/// <summary>
/// Do initial grouping of and ranking of peaks using the Crawdad
/// peak detector.
/// </summary>
public void PickChromatogramPeaks(double[] retentionTimes, bool isAlignedTimes)
{
// Make sure chromatograms are in sorted order
_listChromData.Sort((c1, c2) => c1.Key.CompareTo(c2.Key));
// Mark all optimization chromatograms
MarkOptimizationData();
// if (Math.Round(_listChromData[0].Key.Precursor) == 585)
// Console.WriteLine("Issue");
// First use Crawdad to find the peaks
// If any chromatograms have an associated transition, then only find peaks
// in chromatograms with transitions. It is too confusing to the user to
// score peaks based on chromatograms for hidden transitions.
bool hasDocNode = _listChromData.Any(chromData => chromData.DocNode != null);
_listChromData.ForEach(chromData => chromData.FindPeaks(retentionTimes,
// But only for fragment ions to allow hidden MS1 isotopes to participate
hasDocNode && chromData.Key.Source == ChromSource.fragment));
RemoveProductConflictsByTime(retentionTimes);
// Merge sort all peaks into a single list
IList<ChromDataPeak> allPeaks = SplitMS(MergePeaks());
// Inspect 20 most intense peak regions
var listRank = new List<double>();
Assume.IsTrue(_listPeakSets.Count == 0);
for (int i = 0; i < 20 || retentionTimes.Length > 0; i++)
{
if (allPeaks.Count == 0)
break;
ChromDataPeak peak = allPeaks[0];
allPeaks.RemoveAt(0);
// If peptide ID retention times are present, allow
// peaks greater than 20, but only if they contain
// an ID retention time.
if (i >= 20 && !peak.Peak.Identified)
continue;
ChromDataPeakList peakSet = FindCoelutingPeaks(peak, allPeaks);
peakSet.SetIdentified(retentionTimes, isAlignedTimes);
_listPeakSets.Add(peakSet);
listRank.Add(i);
}
if (_listPeakSets.Count == 0)
return;
// Sort by total area descending
_listPeakSets.Sort((p1, p2) => Comparer<double>.Default.Compare(p2.TotalArea, p1.TotalArea));
// The peak will be a signigificant spike above the norm for this
// data. Find a cut-off by removing peaks until the remaining
// peaks correlate well in a linear regression.
var listAreas = _listPeakSets.ConvertAll(set => set.TotalArea);
// Keep at least 3 peaks
listRank.RemoveRange(0, Math.Min(MINIMUM_PEAKS, listRank.Count));
listAreas.RemoveRange(0, Math.Min(MINIMUM_PEAKS, listAreas.Count));
Assume.IsTrue(listRank.Count == listAreas.Count);
int iRemove = 0;
// Keep all peaks for summary chromatograms
if (PrecursorMz != 0)
{
// And there must be at least 5 peaks in the line to qualify for removal
for (int i = 0, len = listAreas.Count; i < len - 4; i++)
{
var statsRank = new Statistics(listRank);
var statsArea = new Statistics(listAreas);
double rvalue = statsArea.R(statsRank);
// Console.WriteLine("i = {0}, r = {1}", i, rvalue);
if (Math.Abs(rvalue) > NOISE_CORRELATION_THRESHOLD)
{
iRemove = i + MINIMUM_PEAKS;
break;
}
listRank.RemoveAt(0);
listAreas.RemoveAt(0);
}
}
if (iRemove == 0)
{
iRemove = _listPeakSets.Count;
if (retentionTimes.Length == 0)
{
// Backward compatibility: before peptide IDs were integrated
// this sorting happened before peaks were extended.
_listPeakSets.Sort(ComparePeakLists);
}
}
else if (retentionTimes.Length == 0)
{
// Be sure not to remove anything with a higher combined score than
// what happen to look visually like the biggest peaks.
double minKeepScore = _listPeakSets.Take(iRemove).Min(peakSet => peakSet.CombinedScore);
// Backward compatibility: before peptide IDs were integrated
// this sorting happened before peaks were extended.
_listPeakSets.Sort(ComparePeakLists);
iRemove = Math.Max(iRemove, _listPeakSets.IndexOf(peakSet => peakSet.CombinedScore == minKeepScore));
}
else
{
// Make sure no identified peaks are removed.
int identIndex = _listPeakSets.LastIndexOf(peakSet => peakSet.IsIdentified);
if (identIndex >= iRemove)
iRemove = identIndex + 1;
// Or, if there were identifications, but no peaks
// peaks that appear to contain them, keep all peak sets
else if (identIndex != -1)
iRemove = _listPeakSets.Count;
}
RemoveNonOverlappingPeaks(_listPeakSets, iRemove);
// Since Crawdad can have a tendency to pick peaks too narrow,
// use the peak group information to extend the peaks to make
// them wider.
// This does not handle reintegration, because peaks get reintegrated
// before they are stored, taking the entire peptide into account.
_listPeakSets = ExtendPeaks(_listPeakSets, retentionTimes, isAlignedTimes);
// Sort by whether a peak contains an ID and then product score
// This has to be done after peak extending, since extending may
// change the state of whether the peak contains an ID.
if (retentionTimes.Length > 0)
_listPeakSets.Sort(ComparePeakLists);
// if (retentionTimes.Length > 0 && !_listPeakSets[0].Identified)
// Console.WriteLine("Idenifications outside peaks.");
}